Unmanned aerial vehicle flight state classification method and device and storage medium

Abstract

The invention relates to the technical field of intelligent transportation, in particular to an unmanned aerial vehicle flight state classification method, device and equipment and a computer storage medium. According to the invention, the interference of receiver hardware and a complex electromagnetic environment where receiving is carried out in the receiving process of the communication radio frequency signal of the unmanned aerial vehicle is reduced, and the influence of the black box property of a neural network model utilized by the radio frequency-based unmanned aerial vehicle detection and identification method on the state identification accuracy is reduced; a reliable Bayesian causal reasoning enhanced unmanned aerial vehicle target state recognition method is provided for a high-precision recognition theoretical method of a low-altitude three-dimensional traffic unmanned aerial vehicle detection and recognition system in a complex urban environment, and the accuracy of state classification and recognition is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a method, device, equipment and computer storage medium for classifying the flight status of an unmanned aerial vehicle. Background Art

[0002] The development of Internet of Things technology has made urban smart transportation possible. Drones, with their flexibility, adaptability and safety, have generated huge economic and social benefits today. Many countries have or plan to open up low-altitude airspace to provide policy guarantees for drones in logistics distribution, social governance, emergency rescue and other application scenarios. Low-altitude airspace is tightly coupled with ground transportation in the time and space domain, and elements are intertwined in the physical domain to form a low-altitude three-dimensional transportation network. With the vigorous development of low-altitude transportation, the threats and risks that come with it have also emerged. The emergence of malicious drone black flying has seriously affected the normal operation of low-altitude transportation and threatened social public safety. Therefore, it is urgent to establish a fast and effective drone defense system. In this system, the first step is to detect drone targets and accurately identify the drone target status.

[0003] In drone detection and identification systems, there are currently four commonly used methods: acoustic, optical, radar, and radio frequency based methods.

[0004] Compared with acoustic and optical methods, RF detection methods are more stable and reliable. In the urban background, the electromagnetic environment is complex. The use of active radar for detection will affect broadcasting, aircraft communications, etc. RF detection is a passive detection method and will not produce such side effects. Therefore, in the low-altitude three-dimensional traffic network, the drone detection and identification module of the drone defense system is more suitable for using RF-based methods to identify and classify drone patterns. The accuracy of this method for drone pattern recognition is mainly determined by the accuracy of machine learning or neural network classification. In low-altitude three-dimensional traffic, the RF-based drone identification and classification method has the following three challenges. First, in a complex electromagnetic environment, although this method will not affect the communication of other radio equipment, it will be affected by electromagnetic interference from other equipment, and there will be other uncertain factors that affect the accuracy of recognition. Secondly, drones can realize two modes during operation: hovering and flying without image transmission. These two modes of the same drone do not show much difference in the spectrum, which seriously affects the accuracy of recognition. Summary of the invention

[0005] Therefore, the technical problem to be solved by the present invention is to overcome the problem of low accuracy in classification of UAV flight status in the prior art.

[0006] In order to solve the above technical problems, the present invention provides a method for classifying the flight status of a UAV, comprising:

[0007] Collect UAV RF signal data in different scenarios, perform preprocessing and time-frequency conversion to obtain UAV spectrum data;

[0008] The uncertainty factors in the acquisition process of UAV RF signal data are modeled offline, and the distribution of uncertainty factors in the acquisition process of UAV RF signal data is estimated based on the uncertainty factor model;

[0009] According to the influence of the drone on the communication quality under different flight states, combined with the prior knowledge in the field, a causal chain of the drone radio frequency signal is constructed, and the drone radio frequency signal causal chain is trained according to the drone spectrum data and the distribution of the uncertain factors to obtain conditional probability distribution parameters;

[0010] Acquire drone spectrum data in real time and input it into the drone flight status classification neural network model to obtain the drone flight status classification results;

[0011] The flight state classification result of the UAV is corrected based on the UAV radio frequency signal causal chain and the conditional probability distribution parameters to obtain a target classification result.

[0012] Preferably, the collecting of drone radio frequency signal data in different scenarios includes:

[0013] In the preset frequency band, radio signals larger than the preset bandwidth are collected from drones of different manufacturers and models in environments without any drone RF background, when the drone is turned on and connected, when the drone is hovering, when the drone is flying but not transmitting image, and when the drone is flying and transmitting image, to obtain drone RF signal data.

[0014] Preferably, the preprocessing and time-frequency conversion include:

[0015] Performing denoising and decoupling processing on the UAV radio frequency signal data to obtain a desired UAV communication signal;

[0016] The desired UAV communication signal is transformed into a time-frequency signal using discrete Fourier transform, and its phase characteristics are ignored, and only its amplitude characteristics are retained to obtain the UAV spectrum data.

[0017] Preferably, the estimation of the distribution of uncertainties of the drone radio frequency signal data during the acquisition process based on the uncertainty factor model includes:

[0018] Based on the uncertainty factor model, the agent-based modeling and simulation method is used to obtain the initial distribution of uncertainty factors according to domain prior knowledge, and combined with the Monte Carlo simulation method, the distribution of uncertainty factors in the acquisition process of UAV RF signal data is estimated through multiple iterations.

[0019] Preferably, the construction of the causal chain of the UAV radio frequency signal according to the influence of the UAV on the communication quality under different flight states and in combination with the prior knowledge of the domain includes:

[0020] For the UAV radio frequency signal data, the standard deviation and kurtosis are selected as causal nodes in the time domain, and the average value is selected as the causal node in the frequency domain;

[0021] Regarding the communication quality, the communication signal-to-noise ratio, synchronization error and co-frequency interference are selected as causal nodes;

[0022] Combined with domain prior knowledge, causal nodes are connected to build a causal chain of drone RF signals.

[0023] Preferably, the correction of the drone flight state classification result based on the drone radio frequency signal causal chain and the conditional probability distribution parameters to obtain the target classification result includes:

[0024] Based on the causal chain of the UAV radio frequency signal and the conditional probability distribution parameters, a Bayesian network model is used to perform causal reasoning analysis and correction on the UAV flight state classification result.

[0025] Preferably, when the UAV flight state classification neural network model is a multi-channel convolutional neural network model, the UAV spectrum data is evenly divided according to the bandwidth and then input into the channel convolutional neural network model in parallel;

[0026] When the UAV flight state classification neural network model is a long short-term memory neural network model, the UAV spectrum data is normalized and then input into the long short-term memory neural network model;

[0027] When the UAV flight status classification neural network model is a machine learning model, the UAV spectrum data is subjected to feature extraction and then input into the machine learning model.

[0028] The present invention also provides a UAV flight status classification device, comprising:

[0029] The data acquisition module is used to collect the UAV RF signal data in different scenarios, and perform preprocessing and time-frequency conversion to obtain the UAV spectrum data;

[0030] The uncertainty factor acquisition module is used to perform offline modeling on the uncertainty factors in the acquisition process of the UAV RF signal data, and estimate the distribution of the uncertainty factors in the acquisition process of the UAV RF signal data based on the uncertainty factor model;

[0031] A causal chain construction module is used to construct a causal chain of UAV radio frequency signals according to the impact of the UAV on the communication quality under different flight states and in combination with domain prior knowledge, and to train the causal chain of UAV radio frequency signals according to the UAV spectrum data and the distribution of uncertain factors to obtain conditional probability distribution parameters;

[0032] The state classification module is used to obtain the UAV spectrum data in real time and input it into the UAV flight state classification neural network model to obtain the UAV flight state classification result;

[0033] The result correction module is used to correct the UAV flight state classification result based on the UAV radio frequency signal causal chain and the conditional probability distribution parameters to obtain a target classification result.

[0034] The present invention also provides a UAV flight status classification device, comprising:

[0035] Memory for storing computer programs;

[0036] The processor is used to implement the above-mentioned steps of the method for classifying the flight status of a drone when executing the computer program.

[0037] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for classifying the flight status of a drone are implemented.

[0038] The above technical solution of the present invention has the following advantages compared with the prior art:

[0039] The method for classifying the flight state of a UAV described in the present invention first collects and preprocesses spectrum signals to construct a spectrum data set, then selects uncertainty factors and performs offline training to obtain parameter distribution, and constructs a causal chain for flight state classification, performs discretization training on the flight state classification, and finally selects a category that needs further correction based on the neural network classification, and corrects the category according to the constructed causal chain and model parameters; the present invention reduces the interference of the receiver hardware and the complex electromagnetic environment in which the UAV communication radio frequency signal is received during the reception process, and the influence of the black box property of the neural network model used by the radio frequency-based UAV detection and identification method itself on the state recognition accuracy, provides a reliable Bayesian causal reasoning-enhanced UAV target state recognition method for a high-precision recognition theoretical method of a low-altitude three-dimensional traffic UAV detection and identification system in a complex urban environment, and improves the accuracy of state classification recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein:

[0041] Figure 1 This is a flow chart of an implementation of a method for classifying the flight status of a UAV provided by the present invention;

[0042] Figure 2 The five radio frequency backgrounds without drones and the distribution diagram of drone spectrum data in different states provided by the embodiments of the present invention are shown;

[0043] Figure 3 Schematic diagram of the cause-effect chain of the UAV radio frequency signal provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The core of the present invention is to provide a method, device, equipment and computer storage medium for classifying the flight status of an unmanned aerial vehicle, which effectively improves the accuracy of classification and identification.

[0045] In order to enable those skilled in the art to better understand the scheme of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0046] Please refer to Figure 1 , Figure 1 This is a flowchart of the implementation of a UAV flight status classification method provided by the present invention; the specific operation steps are as follows:

[0047] S101: Collect UAV radio frequency signal data in different scenarios, and perform preprocessing and time-frequency conversion to obtain UAV spectrum data;

[0048] S102: Offline modeling of uncertain factors in the collection process of the UAV radio frequency signal data, and estimating the distribution of uncertain factors in the collection process of the UAV radio frequency signal data based on the uncertainty factor model;

[0049] S103: According to the influence of the drone on the communication quality under different flight states, combined with the prior knowledge in the field, a causal chain of the drone radio frequency signal is constructed, and the drone radio frequency signal causal chain is trained according to the drone spectrum data and the distribution of the uncertain factors to obtain conditional probability distribution parameters;

[0050] S104: Acquire the drone spectrum data in real time, and input it into the drone flight state classification neural network model to obtain the drone flight state classification result;

[0051] S105: Based on the causal chain of the UAV radio frequency signal and the conditional probability distribution parameters, the UAV flight state classification result is corrected to obtain a target classification result.

[0052] Based on the above embodiments, this embodiment describes step S101 in detail:

[0053] The acquisition of drone radio frequency signal data in different scenarios includes:

[0054] In the preset frequency band, radio signals larger than the preset bandwidth are collected from drones of different manufacturers and models in environments without any drone RF background, when the drone is turned on and connected, when the drone is hovering, when the drone is flying but not transmitting image, and when the drone is flying and transmitting image, to obtain drone RF signal data.

[0055] The present invention requires the use of an acquisition system and a data processing system to receive and preprocess the radio frequency data of the drone. In one embodiment, a radio receiving device, such as a USRP series software-defined radio peripheral, is used to collect radio signals with a bandwidth of at least 80MHz in the 2.4GHz or 5.8GHz frequency band for multiple scenarios of non-existent radio frequency background environments of drones of different manufacturers and models, drones being turned on and connected, drones hovering, drones flying but not transmitting images, and drones flying and transmitting images. The commonly used communication frequency bands for drones are the 2.4GHz or 5.8GHz frequency bands. Therefore, when selecting a radio receiving device, the receivable frequency range must be considered to ensure that the device can collect data in this communication frequency band. To collect UAV communication RF data, at least 80MHz bandwidth of data is required to be suitable for identification and analysis. Therefore, when selecting a radio receiving device, it is necessary to consider not only the receivable frequency range, but also the maximum receiving bandwidth of the radio receiving device. If the maximum receiving bandwidth cannot meet the needs, multiple radio receiving devices can be used for data collection at the same time, but a time synchronization module needs to be installed on each device to basically ensure that the collected data is collected at the same time. In this way, the collected data can be spliced ​​in the frequency domain, and a RF communication signal data set with at least 80MHz bandwidth can also be obtained.

[0056] The preprocessing and time-frequency conversion include:

[0057] The UAV radio frequency signal data is denoised and decoupled to obtain the desired UAV communication signal. In one embodiment, an underdetermined blind source separation method based on sparse variable analysis is used, and the time domain signal is first converted to the time-frequency domain through time-frequency transform such as wavelet transform, short-time Fourier transform, etc., and a single source point is obtained using a data analysis method under a specific frequency window. Then, a clustering operation is performed on the single source point to obtain a mixing matrix, and finally, the desired UAV communication signal and other noise signals are reconstructed to achieve denoising and decoupling of the perception signal.

[0058] The desired drone communication signal is transformed into a time-frequency signal using a discrete Fourier transform, and its phase characteristics are ignored, and only its amplitude characteristics are retained to obtain the drone spectrum data. In one embodiment, assuming that the number of frequency windows is set to 2048, then the frequency domain data after the discrete Fourier transform has a total of 2048 frequency windows, that is, each drone communication power spectrum data is reduced from tens of thousands of data points in the time domain to 2048 data points in the frequency domain. The distribution diagrams of the five types of RF backgrounds without drones and drone spectrum data in different states are shown in Figure 1. Figure 2 shown.

[0059] Based on the above embodiments, this embodiment describes step S102 in detail:

[0060] In the offline state of the low-altitude UAV detection and identification system, a model of uncertainty factors from the environment and acquisition equipment that may be affected by the drone spectrum data during the acquisition process is designed. When the maximum receiving bandwidth of the selected radio acquisition device is not enough to cover the bandwidth required for the drone RF data analysis, it is necessary to adopt a method of simultaneous acquisition of multiple devices. Although the synchronization module can be used to ensure time synchronization when performing multi-device acquisition, there is still a problem of synchronization error. And in the acquisition process, since the 2.4GHz frequency band is used by WIFI, Bluetooth, etc. in addition to drone communication signals, there will be a problem of co-channel interference during the acquisition process. Therefore, when facing the collected drone RF data, it is essential to consider uncertainty factors. For example, considering uncertainty factors such as synchronization error and co-channel interference will affect the accuracy of RF-based drone target state recognition;

[0061] Using the agent-based modeling and simulation method, the initial distribution of the proposed uncertainty factors is given according to the domain prior knowledge. The Monte Carlo simulation method is used through multiple iterations to obtain the distribution of uncertainty factors that may be encountered in the collection of UAV communication RF data.

[0062] Based on the above embodiments, this embodiment describes step S103 in detail:

[0063] Causal nodes are selected based on the impact of drones on communication quality under different flight conditions. In one embodiment, for drone communication RF signals, standard deviation and kurtosis are selected as causal nodes in the time domain, and average value is selected as causal node in the frequency domain. For communication quality, communication signal-to-noise ratio is selected as causal node, and two uncertain factors, synchronization error and co-frequency interference, are selected as causal nodes. Combined with domain prior knowledge, causal nodes are connected to construct a causal chain of drone RF signals. The constructed causal chain of drone RF signals is as follows: Figure 3 As shown in Figure 1, the causal chain is a directed acyclic graph consisting of nodes and directed edges connecting the nodes. Each node represents a random variable, which can be an observed variable, a latent variable, or an unknown parameter, and the directed edges between nodes represent the causal relationship between the variables.

[0064] The causal chain constructed based on the acquired radio frequency signal and the sampling uncertainty factors is trained to obtain the conditional probability distribution parameters.

[0065] Based on the above embodiments, this embodiment describes step S104 in detail:

[0066] The UAV RF power spectrum data is processed according to the input requirements of different neural network models. Assuming that the input neural network model is a multi-channel convolutional neural network, the UAV communication power spectrum data with an 80MHz bandwidth needs to be evenly divided into 8 independent 10MHz bandwidth segments, that is, there are 256 data points in each bandwidth segment corresponding to each UAV power spectrum data, and each bandwidth segment is input into the 8-channel convolutional neural network in parallel as a channel. Assuming that the input neural network model is a long short-term memory neural network, the power spectrum only needs to be simply normalized. Assuming that the input is a machine learning model, the features of the power spectrum data need to be manually extracted and then input into the machine learning model. The processed power spectrum data is input into the neural network model for feature extraction, feature dimension reduction and target classification to obtain the initial low-altitude UAV flight status classification results.

[0067] Based on the above embodiments, this embodiment describes step S105 in detail:

[0068] For the initial low-altitude UAV flight status classification results, select the categories that need further correction; for the situation where different states of the same UAV do not show much difference in the spectrum, the neural network does not achieve good feature extraction and classification recognition, and the opacity of the neural network model as a black box model is often the cause of such false positives and false negatives. For these black box model recognition errors, the Bayesian network model is used to perform causal reasoning analysis and correction, increase the transparency of the recognition model, enhance the interpretability of the recognition model, and thus improve the accuracy of UAV target state recognition. Combined with the aforementioned causal chain constructed and the probability distribution parameters obtained through training, the model results of the neural network training are corrected to output the final results of the urban low-altitude UAV flight status classification.

[0069] In summary, the present invention can be used to detect the spectrum of unmanned aerial vehicles (UAVs) based on UAVs, and can achieve high-precision recognition of the target state of UAVs according to actual needs, reduce the interference of the receiver hardware and the complex electromagnetic environment in which the UAV communication RF signal is received during the reception process, and the influence of the black box nature of the neural network model on the accuracy of state recognition by the RF-based UAV detection and identification method itself. It provides a reliable Bayesian causal reasoning-enhanced UAV target state recognition method for a high-precision recognition theoretical method of a low-altitude three-dimensional traffic UAV detection and identification system in a complex urban environment, and has strong applicability.

[0070] The embodiment of the present invention also provides a device for classifying the flight status of a drone; the specific device may include:

[0071] The data acquisition module is used to collect the UAV RF signal data in different scenarios, and perform preprocessing and time-frequency conversion to obtain the UAV spectrum data;

[0072] The uncertainty factor acquisition module is used to perform offline modeling on the uncertainty factors in the acquisition process of the UAV RF signal data, and estimate the distribution of the uncertainty factors in the acquisition process of the UAV RF signal data based on the uncertainty factor model;

[0073] A causal chain construction module is used to construct a causal chain of UAV radio frequency signals according to the impact of the UAV on the communication quality under different flight states and in combination with domain prior knowledge, and to train the causal chain of UAV radio frequency signals according to the UAV spectrum data and the distribution of uncertain factors to obtain conditional probability distribution parameters;

[0074] The state classification module is used to obtain the UAV spectrum data in real time and input it into the UAV flight state classification neural network model to obtain the UAV flight state classification result;

[0075] The result correction module is used to correct the UAV flight state classification result based on the UAV radio frequency signal causal chain and the conditional probability distribution parameters to obtain a target classification result.

[0076] The drone flight state classification device of this embodiment is used to implement the aforementioned drone flight state classification method. Therefore, the specific implementation method of the drone flight state classification device can be seen in the implementation example part of the drone flight state classification method mentioned above. For example, the data acquisition module, the uncertainty factor acquisition module, the causal chain construction module, the state classification module, and the result correction module are respectively used to implement steps S101, S102, S103, S104 and S105 in the above-mentioned drone flight state classification method. Therefore, its specific implementation method can refer to the description of the corresponding each part of the implementation example, which will not be repeated here.

[0077] A specific embodiment of the present invention further provides a drone flight status classification device, comprising: a memory for storing a computer program; and a processor for implementing the steps of the above-mentioned drone flight status classification method when executing the computer program.

[0078] A specific embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for classifying the flight status of a drone are implemented.

[0079] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0080] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0081] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0082] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0083] Obviously, the above embodiments are merely examples for the purpose of clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the present invention.

Claims

1. A method for classifying the flight status of a drone, characterized in that: include: Collect UAV RF signal data in different scenarios, perform preprocessing and time-frequency conversion to obtain UAV spectrum data; The uncertainty factors in the acquisition process of UAV RF signal data are modeled offline, and the distribution of uncertainty factors in the acquisition process of UAV RF signal data is estimated based on the uncertainty factor model; According to the impact of drones on communication quality in different flight states, combined with prior knowledge in the field, Constructing a causal chain of UAV radio frequency signals, and training the causal chain of UAV radio frequency signals according to the UAV spectrum data and the distribution of uncertain factors to obtain conditional probability distribution parameters; Acquire drone spectrum data in real time and input it into the drone flight status classification neural network model to obtain the drone flight status classification results; The flight state classification result of the UAV is corrected based on the UAV radio frequency signal causal chain and the conditional probability distribution parameters to obtain a target classification result.

2. The method for classifying the flight status of a UAV according to claim 1, characterized in that: The acquisition of drone radio frequency signal data in different scenarios includes: In the preset frequency band, radio signals larger than the preset bandwidth are collected from drones of different manufacturers and models in environments without any drone RF background, when the drone is turned on and connected, when the drone is hovering, when the drone is flying but not transmitting image, and when the drone is flying and transmitting image, to obtain drone RF signal data.

3. The method for classifying the flight status of a UAV according to claim 1, characterized in that: The preprocessing and time-frequency conversion include: Performing denoising and decoupling processing on the UAV radio frequency signal data to obtain a desired UAV communication signal; The desired UAV communication signal is transformed into a time-frequency signal using discrete Fourier transform, and its phase characteristics are ignored, and only its amplitude characteristics are retained to obtain the UAV spectrum data.

4. The method for classifying the flight status of a UAV according to claim 1, characterized in that: The uncertainty factor distribution of the drone radio frequency signal data during the acquisition process estimated based on the uncertainty factor model includes: Based on the uncertainty factor model, the agent-based modeling and simulation method is used to obtain the initial distribution of uncertainty factors according to domain prior knowledge, and combined with the Monte Carlo simulation method, the distribution of uncertainty factors in the acquisition process of UAV RF signal data is estimated through multiple iterations.

5. The method for classifying the flight status of a UAV according to claim 1, characterized in that: According to the influence of drones on communication quality under different flight conditions and combined with prior knowledge in the field, the causal chain of drone radio frequency signals is constructed, including: For the UAV radio frequency signal data, the standard deviation and kurtosis are selected as causal nodes in the time domain, and the average value is selected as the causal node in the frequency domain; Regarding the communication quality, the communication signal-to-noise ratio, synchronization error and co-frequency interference are selected as causal nodes; Combined with domain prior knowledge, causal nodes are connected to build a causal chain of drone RF signals.

6. The method for classifying the flight status of a UAV according to claim 1, characterized in that: The correcting of the UAV flight state classification result based on the UAV radio frequency signal causal chain and the conditional probability distribution parameters to obtain the target classification result includes: Based on the causal chain of the UAV radio frequency signal and the conditional probability distribution parameters, a Bayesian network model is used to perform causal reasoning analysis and correction on the UAV flight state classification result.

7. The method for classifying the flight status of a UAV according to claim 1, characterized in that: When the UAV flight state classification neural network model is a multi-channel convolutional neural network model, the UAV spectrum data is evenly divided according to the bandwidth and then input into the channel convolutional neural network model in parallel; When the UAV flight state classification neural network model is a long short-term memory neural network model, the UAV spectrum data is normalized and then input into the long short-term memory neural network model; When the UAV flight status classification neural network model is a machine learning model, the UAV spectrum data is subjected to feature extraction and then input into the machine learning model.

8. A drone flight status classification device, characterized in that: include: The data acquisition module is used to collect the UAV RF signal data in different scenarios, and perform preprocessing and time-frequency conversion to obtain the UAV spectrum data; The uncertainty factor acquisition module is used to perform offline modeling on the uncertainty factors in the acquisition process of the UAV RF signal data, and estimate the distribution of the uncertainty factors in the acquisition process of the UAV RF signal data based on the uncertainty factor model; A causal chain construction module is used to construct a causal chain of UAV radio frequency signals according to the impact of the UAV on the communication quality under different flight states and in combination with domain prior knowledge, and to train the causal chain of UAV radio frequency signals according to the UAV spectrum data and the distribution of uncertain factors to obtain conditional probability distribution parameters; The state classification module is used to obtain the UAV spectrum data in real time and input it into the UAV flight state classification neural network model to obtain the UAV flight state classification result; The result correction module is used to correct the UAV flight state classification result based on the UAV radio frequency signal causal chain and the conditional probability distribution parameters to obtain a target classification result.

9. A drone flight status classification device, characterized in that: include: Memory for storing computer programs; A processor, used to implement the steps of a method for classifying the flight status of a drone as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for classifying the flight status of a drone as described in any one of claims 1 to 7 are implemented.

Images