An unmanned aerial vehicle intelligent network exploration and inspection method and system

Abstract

The application discloses an unmanned aerial vehicle intelligent network surveying and inspecting method and system, and relates to the technical field of data analysis.The method comprises the following steps: an unmanned aerial vehicle collects full-band spectrum data, antenna transmitting power information and base station antenna image pairs for preprocessing; based on the full-band spectrum data, pure main lobe radiation power spectrum analysis is carried out to perform antenna transmitting point attribution analysis; based on the antenna transmitting power information, antenna transmitting power anomaly detection is performed; based on the base station antenna image, antenna contour regions are analyzed to segment antenna dipoles, and based on the antenna dipoles, antenna direction angle and downtilt angle attitude information are identified; based on the antenna transmitting point attribution information, antenna transmitting power anomaly detection results, antenna direction angle and downtilt angle attitude information, network security hidden danger surveying is performed to generate a surveying and inspecting report.The application converts scattered data into operation files which can directly guide network optimization and operation and maintenance decisions, and realizes more reliable unmanned aerial vehicle intelligent network surveying and inspecting.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a method and system for intelligent network surveying and inspection using unmanned aerial vehicles (UAVs). Background Technology

[0002] With the development of UAV technology, UAVs have been gradually applied to the survey and inspection of base station network systems. However, existing UAV survey and inspection solutions typically only acquire the overall spectrum waveform in spectrum measurement, failing to extract the pure main lobe radiation power spectrum and lacking the ability to analyze antenna transmission point attribution using this power spectrum. This makes it difficult to accurately identify the antenna source of the radiated signal in multi-antenna co-location scenarios. Regarding antenna attitude detection, existing methods mainly rely on visual interpretation of acquired images by relevant personnel, without considering the analysis of the antenna contour region or antenna element segmentation. This results in the inability to effectively identify the azimuth and downtilt attitudes, making it difficult to guarantee measurement accuracy and consistency. Furthermore, current verification of antenna transmission power still relies primarily on manual comparison, lacking the means to detect anomalies based on power information, making it easy for hidden power faults to go undetected in a timely manner. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a method and system for intelligent network surveying and inspection by unmanned aerial vehicles (UAVs), which transforms scattered data into operational documents that can directly guide network optimization and maintenance decisions, thus realizing the value of intelligent network surveying and inspection by UAVs.

[0004] To address the aforementioned technical problems, this invention provides a method for intelligent network surveying and inspection using unmanned aerial vehicles (UAVs), the method comprising: The UAV carries a lightweight frequency measurement payload to collect full-band spectrum data, antenna transmit power information, and base station antenna images. It also preprocesses the full-band spectrum data, antenna transmit power information, and base station antenna images to obtain preprocessed full-band spectrum data, antenna transmit power information, and base station antenna images. Based on the preprocessed full-band spectrum data, a pure main lobe radiated power spectrum analysis is performed to obtain pure main lobe radiated power spectrum information. Based on the pure main lobe radiated power spectrum information, an antenna transmission point attribution analysis is performed to obtain antenna transmission point attribution information. Antenna transmit power anomaly detection is performed based on the preprocessed antenna transmit power information to obtain antenna transmit power anomaly detection results; The antenna contour region is analyzed based on the preprocessed base station antenna image, and the antenna element is segmented based on the antenna contour region. Attitude recognition is performed based on the antenna element to obtain the antenna azimuth angle and downtilt angle attitude information. Based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth angle and downtilt angle attitude information, a network security vulnerability investigation is conducted to obtain network security vulnerability investigation results, and an investigation and inspection report is generated based on the network security vulnerability investigation results.

[0005] Optionally, the preprocessing of the full-band spectrum data, antenna transmit power information, and base station antenna image to obtain preprocessed full-band spectrum data, antenna transmit power information, and base station antenna image includes: Harmonic suppression is performed on the full-band spectrum data to obtain harmonic-suppressed full-band spectrum data, and multi-scale wavelet packet decomposition is performed on the harmonic-suppressed full-band spectrum data to obtain preprocessed full-band spectrum data. The antenna transmit power information is cleaned to obtain preprocessed antenna transmit power information; The base station antenna image is dehazed to obtain a dehazed base station antenna image, and then geometric correction is performed on the dehazed base station antenna image to obtain a preprocessed base station antenna image.

[0006] Optionally, the step of performing pure main lobe radiated power spectrum analysis based on preprocessed full-band spectrum data to obtain pure main lobe radiated power spectrum information, and performing antenna transmission point attribution analysis based on the pure main lobe radiated power spectrum information to obtain antenna transmission point attribution information, includes: A high-resolution time-frequency spectrum is determined based on the preprocessed full-band spectrum data, and a clean main lobe radiation power spectrum analysis is performed based on the high-resolution time-frequency spectrum to obtain clean main lobe radiation power spectrum information. Based on the pure main lobe radiation power spectrum information, candidate legitimate base station identities are screened to obtain a set of candidate legitimate base station identities. Based on the set of candidate legitimate base station identities, perform posterior probability analysis on each candidate base station to obtain the corresponding posterior probability information; Antenna transmission point attribution analysis is performed based on the posterior probability information to obtain antenna transmission point attribution information.

[0007] Optionally, the step of performing pure mainlobe radiation power spectrum analysis based on the high-resolution time-spectrum to obtain pure mainlobe radiation power spectrum information includes: The high-resolution time-frequency spectrum is denoised to obtain the denoised high-resolution time-frequency spectrum. The main lobe spectral contour is extracted based on the high-resolution time-frequency spectrum after noise reduction processing to obtain the main lobe spectral contour information. Based on the main lobe spectral contour information, the spectral identifier of the standard frame structure signal is determined, and time-frequency mask analysis is performed based on the spectral identifier to obtain the target time-frequency mask; Based on the target time-frequency mask, a pure main lobe radiation power spectrum analysis is performed to obtain pure main lobe radiation power spectrum information.

[0008] Optionally, determining the spectral identifier of the canonical frame structure signal based on the main lobe spectral contour information includes: Based on the main lobe spectral contour information, a two-dimensional cyclic spectral function graph is determined, and based on the two-dimensional cyclic spectral function graph, the cyclic stationary characteristic peak analysis of the carrier signal is performed to obtain cyclic stationary characteristic peak information; Based on the cyclic stationary characteristic peak information, cross-correlation matching is performed using a preset cyclic stationary characteristic peak template library to obtain cross-correlation matching results. Based on the cross-correlation matching results, the spectral identifier of the standard frame structure signal is determined using a preset decision threshold.

[0009] Optionally, the step of detecting antenna transmit power anomalies based on preprocessed antenna transmit power information to obtain antenna transmit power anomaly detection results includes: Short-time power mean analysis is performed based on the preprocessed antenna transmit power information to obtain a short-time power mean sequence. The long-term autocorrelation attenuation coefficient sequence is determined based on the preprocessed antenna transmit power information, and the target feature vector is determined based on the short-term power mean sequence and the long-term autocorrelation attenuation coefficient sequence. A reference power curve is obtained, and antenna transmit power anomaly detection is performed based on the reference power curve and the target feature vector to obtain the antenna transmit power anomaly detection result.

[0010] Optionally, the step of detecting antenna transmit power anomalies based on the reference power curve and the target feature vector to obtain antenna transmit power anomaly detection results includes: Based on the reference power curve, antenna transmit power offset analysis is performed using the short-time power mean sequence to obtain the antenna transmit power offset sequence. Anomaly marking analysis is performed based on the antenna transmit power offset sequence and target feature vector to obtain anomaly marking sequence; Antenna transmit power anomaly detection is performed based on the anomaly marker sequence to obtain antenna transmit power anomaly detection results.

[0011] Optionally, the step of analyzing the antenna contour region based on the preprocessed base station antenna image, segmenting the antenna element based on the antenna contour region, and performing attitude recognition based on the antenna element to obtain antenna azimuth and downtilt attitude information includes: Antenna edge map extraction is performed based on the preprocessed base station antenna image to obtain the target antenna edge map, and the antenna contour region is determined based on the target antenna edge map; The antenna element features are analyzed based on the antenna contour region to identify the regular texture differences between the antenna elements, and the antenna elements are segmented based on the feature maps. The spatial geometric description of the antenna element is obtained, and attitude recognition is performed based on the spatial geometric description to obtain the antenna orientation angle and downtilt angle attitude information.

[0012] Optionally, the network security vulnerability investigation based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth and downtilt attitude information, to obtain network security vulnerability investigation results, includes: A multi-source heterogeneous dataset is generated based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna orientation angle and downtilt angle attitude information. Based on the aforementioned multi-source heterogeneous dataset, an investigation of potential illegal intrusion risks into base station signals is conducted, and the results of the investigation are obtained. Based on the multi-source heterogeneous dataset, a cross-area coverage interference risk investigation is conducted to obtain the cross-area coverage interference risk investigation results. Based on the illegal intrusion risk investigation results and the cross-area coverage interference risk investigation results, the network security risk investigation results are determined.

[0013] In addition, the present invention also provides an intelligent network survey and inspection system for unmanned aerial vehicles (UAVs), the system comprising: Preprocessing module: Used by UAV carrying a lightweight frequency measurement payload to collect full-band spectrum data, antenna transmit power information and base station antenna images, and to preprocess the full-band spectrum data, antenna transmit power information and base station antenna images to obtain preprocessed full-band spectrum data, antenna transmit power information and base station antenna images; Transmitter point attribution module: used to perform pure main lobe radiated power spectrum analysis based on preprocessed full-band spectrum data to obtain pure main lobe radiated power spectrum information, and to perform antenna transmitter point attribution analysis based on the pure main lobe radiated power spectrum information to obtain antenna transmitter point attribution information; Anomaly detection module: used to detect antenna transmit power anomalies based on preprocessed antenna transmit power information and obtain antenna transmit power anomaly detection results; Attitude recognition module: used to analyze the antenna contour region based on the preprocessed base station antenna image, segment the antenna element based on the antenna contour region, and perform attitude recognition based on the antenna element to obtain antenna azimuth angle and downtilt angle attitude information; Report generation module: used to conduct network security vulnerability investigation based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, antenna azimuth angle and downtilt angle attitude information, obtain network security vulnerability investigation results, and generate an investigation and inspection report based on the network security vulnerability investigation results.

[0014] In this embodiment of the invention, a UAV equipped with a lightweight frequency measurement payload collects full-band spectrum data, antenna transmit power information, and base station antenna images for preprocessing. Based on the preprocessed full-band spectrum data, a clean main lobe radiated power spectrum analysis is performed. Antenna transmission point attribution analysis is then performed based on the clean main lobe radiated power spectrum information to prevent mismatching of side lobes or reflected signals to incorrect antennas, accurately identifying the antenna source of radiated signals. Antenna transmit power anomaly detection is performed based on the preprocessed antenna transmit power information, eliminating the heavy workload of manual power data verification and improving the accuracy of fault identification. Antenna contour regions are analyzed based on the preprocessed base station antenna images, and antenna elements are segmented based on these regions. Attitude recognition is performed based on the antenna elements to obtain antenna azimuth and downtilt attitude information, ensuring the accuracy and consistency of attitude recognition. Network security vulnerability investigation is conducted based on antenna transmission point attribution information, antenna transmit power anomaly detection results, and antenna azimuth and downtilt attitude information. An investigation and inspection report is generated based on the network security vulnerability investigation results, transforming scattered data into operational documents that can directly guide network optimization and maintenance decisions, realizing the value of UAV-based intelligent network investigation and inspection. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the UAV intelligent network survey and inspection method in an embodiment of the present invention. Figure 2 This is a flowchart illustrating an intelligent network survey and inspection method for unmanned aerial vehicles (UAVs) according to another embodiment of the present invention. Figure 3 This is a schematic diagram of the structural composition of the UAV intelligent network survey and inspection system in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1 Please see Figure 1 , Figure 1This is a flowchart illustrating the UAV intelligent network survey and inspection method according to an embodiment of the present invention. The method includes: S11: The UAV carries a lightweight frequency measurement payload to collect full-band spectrum data, antenna transmit power information and base station antenna images, and preprocesses the full-band spectrum data, antenna transmit power information and base station antenna images to obtain preprocessed full-band spectrum data, antenna transmit power information and base station antenna images; In the specific implementation of this invention, a UAV carrying a lightweight frequency measurement payload collects full-band spectrum data, antenna transmit power information, and base station antenna images. Harmonic suppression is applied to the full-band spectrum data to obtain harmonic-suppressed full-band spectrum data. Multi-scale wavelet packet decomposition is then performed on the harmonic-suppressed full-band spectrum data to obtain preprocessed full-band spectrum data. Data cleaning processing is performed on the antenna transmit power information to obtain preprocessed antenna transmit power information. Dehazing processing is performed on the base station antenna images to obtain dehazed base station antenna images. Geometric correction processing is then performed on the dehazed base station antenna images to obtain preprocessed base station antenna images, laying a reliable data foundation for subsequent refined analysis.

[0019] S12: Perform pure main lobe radiation power spectrum analysis based on the preprocessed full-band spectrum data to obtain pure main lobe radiation power spectrum information, and perform antenna transmission point attribution analysis based on the pure main lobe radiation power spectrum information to obtain antenna transmission point attribution information. In the specific implementation of this invention, a high-resolution time-frequency spectrum is determined based on the preprocessed full-band spectrum data, and a clean main lobe radiated power spectrum analysis is performed based on the high-resolution time-frequency spectrum to obtain clean main lobe radiated power spectrum information. Based on the clean main lobe radiated power spectrum information, candidate legitimate base station identities are screened to obtain a set of candidate legitimate base station identities. Based on the set of candidate legitimate base station identities, posterior probability analysis is performed on each candidate base station to obtain corresponding posterior probability information. Based on the posterior probability information, antenna transmission point attribution analysis is performed to obtain antenna transmission point attribution information, accurately resolving the issue of signal ownership and preventing the mismatch of sidelobes or reflected signals to incorrect antennas, thus providing precise directional basis for coverage assessment and interference localization.

[0020] S13: Detect antenna transmit power anomalies based on the preprocessed antenna transmit power information and obtain antenna transmit power anomaly detection results; In the specific implementation of this invention, short-time power mean analysis is performed based on the preprocessed antenna transmit power information to obtain a short-time power mean sequence; a long-time autocorrelation attenuation coefficient sequence is determined based on the preprocessed antenna transmit power information, and a target feature vector is determined based on the short-time power mean sequence and the long-time autocorrelation attenuation coefficient sequence; a reference power curve is obtained, and antenna transmit power anomaly detection is performed based on the reference power curve and the target feature vector to obtain the antenna transmit power anomaly detection result. This eliminates the high-load work of manually checking power data station by station, significantly improving the timeliness and accuracy of fault identification.

[0021] S14: Analyze the antenna contour region based on the preprocessed base station antenna image, segment the antenna element based on the antenna contour region, and perform attitude recognition based on the antenna element to obtain antenna azimuth angle and downtilt angle attitude information; In the specific implementation of this invention, antenna edge maps are extracted based on preprocessed base station antenna images to obtain target antenna edge maps, and antenna contour regions are determined based on the target antenna edge maps; feature maps of regular texture differences between antenna element units are analyzed based on the antenna contour regions, and antenna elements are segmented based on the feature maps; spatial geometric descriptions of the antenna elements are obtained, and attitude recognition is performed based on the spatial geometric descriptions to obtain antenna azimuth and downtilt attitude information. High-precision automated measurement of antenna attitude is achieved in a non-contact, long-distance manner, replacing dangerous manual tower climbing measurement. The measurement results are traceable and highly verifiable.

[0022] S15: Based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth angle and downtilt angle attitude information, conduct a network security vulnerability survey, obtain network security vulnerability survey results, and generate a survey and inspection report based on the network security vulnerability survey results.

[0023] In the specific implementation of this invention, a multi-source heterogeneous dataset is generated based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth angle and downtilt angle attitude information; an illegal intrusion risk investigation of base station signals is conducted based on the multi-source heterogeneous dataset to obtain illegal intrusion risk investigation results; an over-coverage interference risk investigation is conducted based on the multi-source heterogeneous dataset to obtain over-coverage interference risk investigation results; and a network security risk investigation result is determined based on the illegal intrusion risk investigation result and the over-coverage interference risk investigation result; and an investigation and inspection report is generated based on the network security risk investigation result, thus completing a fully automatic closed loop from data acquisition, feature extraction, anomaly identification to risk location and report output.

[0024] In this embodiment of the invention, a UAV equipped with a lightweight frequency measurement payload collects full-band spectrum data, antenna transmit power information, and base station antenna images for preprocessing. Based on the preprocessed full-band spectrum data, a clean main lobe radiated power spectrum analysis is performed. Antenna transmission point attribution analysis is then performed based on the clean main lobe radiated power spectrum information to prevent mismatching of side lobes or reflected signals to incorrect antennas, accurately identifying the antenna source of radiated signals. Antenna transmit power anomaly detection is performed based on the preprocessed antenna transmit power information, eliminating the heavy workload of manual power data verification and improving the accuracy of fault identification. Antenna contour regions are analyzed based on the preprocessed base station antenna images, and antenna elements are segmented based on these regions. Attitude recognition is performed based on the antenna elements to obtain antenna azimuth and downtilt attitude information, ensuring the accuracy and consistency of attitude recognition. Network security vulnerability investigation is conducted based on antenna transmission point attribution information, antenna transmit power anomaly detection results, and antenna azimuth and downtilt attitude information. An investigation and inspection report is generated based on the network security vulnerability investigation results, transforming scattered data into operational documents that can directly guide network optimization and maintenance decisions, realizing the value of UAV-based intelligent network investigation and inspection.

[0025] Example 2 Please see Figure 2 , Figure 2 This is a flowchart illustrating an intelligent network survey and inspection method for unmanned aerial vehicles (UAVs) according to another embodiment of the present invention. The method includes: S201: The UAV carries a lightweight frequency measurement payload to collect full-band spectrum data, antenna transmit power information and base station antenna images, and preprocesses the full-band spectrum data, antenna transmit power information and base station antenna images to obtain preprocessed full-band spectrum data, antenna transmit power information and base station antenna images; In the specific implementation of this invention, the preprocessing of the full-band spectrum data, antenna transmit power information, and base station antenna image to obtain preprocessed full-band spectrum data, antenna transmit power information, and base station antenna image includes: performing harmonic suppression on the full-band spectrum data to obtain harmonic-suppressed full-band spectrum data, and performing multi-scale wavelet packet decomposition on the harmonic-suppressed full-band spectrum data to obtain preprocessed full-band spectrum data; performing data cleaning processing on the antenna transmit power information to obtain preprocessed antenna transmit power information; performing dehazing processing on the base station antenna image to obtain a dehazed base station antenna image, and performing geometric correction processing on the dehazed base station antenna image to obtain a preprocessed base station antenna image.

[0026] Specifically, a drone equipped with a lightweight frequency measurement payload collects full-band spectrum data, antenna transmit power information, and base station antenna images. This demonstrates the promising large-scale application prospects of lightweight drone payloads. A Raspberry Pi development solution was used to develop a lightweight frequency measurement payload (frequency bandwidth: 700MHz to 6GHz). The payload flies along a three-dimensional spiral path covering the front and sides of a base station sector. When the onboard differential positioning module determines that the distance between the drone and the target base station has entered a preset threshold, it simultaneously triggers full-band spectrum acquisition, antenna aperture transmit power envelope sampling, and multi-angle sequence imaging of the base station antenna. All data are stamped with a high-precision timestamp corrected by satellite second pulses, obtaining full-band spectrum data, antenna transmit power information, and original multi-angle sequence images of the base station antenna.

[0027] Harmonic suppression is performed on the full-band spectrum data. Using the three-axis angular velocity and acceleration of the UAV body recorded by the inertial measurement unit, the fundamental frequency and harmonic frequency array corresponding to the rotor speed are extracted by fast Fourier transform to obtain a dynamic interference spectrum template. Based on the dynamic interference spectrum template, a cascaded second-order notch filter bank with the center frequency adjusted in real time with the rotation speed is constructed. The full-band spectrum data is subjected to frame-by-frame notch filtering to effectively remove the narrowband strong interference introduced by the UAV rotor motor, and the harmonic-suppressed full-band spectrum data is obtained. Multi-scale wavelet packet decomposition is performed on the harmonic-suppressed full-band spectrum data. Soft thresholds are calculated on each decomposed sub-band based on the general threshold rule, and the wavelet coefficients are shrunk. The processed coefficients are reconstructed into time-domain signals, thereby achieving the separation of wideband noise floor and random impulse interference, and obtaining preprocessed full-band spectrum data.

[0028] The antenna transmit power information is cleaned and processed. The peak holding voltage is extracted from the antenna transmit power information at millisecond intervals. It is converted into an initial received power value through the load factory calibration curve and synchronously associated with the three-dimensional position of the UAV and the coordinates of the base station antenna at this moment to obtain an initial received power sequence with spatial tags. Using the initial received power sequence, the propagation loss is calculated by substituting it into the free space path loss model based on the real-time straight-line distance between the UAV and the base station antenna. The gain value of the antenna at the corresponding frequency band and incident angle is then replenished and divided by the nominal directional gain of the base station antenna to deduce the transmit power equivalent to the antenna aperture, thus obtaining the preprocessed antenna transmit power information.

[0029] The base station antenna image is dehazed, the dark channel prior information of each frame is extracted, the global atmospheric light value is estimated and the transmittance map is calculated, and the transmittance edge is refined by guided filtering to restore a clear and fog-free base station antenna image. The dehazed base station antenna image is then geometrically corrected. Scale-invariant feature points are detected between adjacent frames and nearest neighbor matching is performed. The homography matrix is ​​estimated using a random sampling consistency algorithm to correct perspective distortion and rotational misalignment caused by UAV hovering micro-motions and airflow disturbances, thus obtaining the preprocessed base station antenna image.

[0030] S202: Determine the high-resolution time spectrum based on the preprocessed full-band spectrum data, and perform pure main lobe radiation power spectrum analysis based on the high-resolution time spectrum to obtain pure main lobe radiation power spectrum information; In a specific implementation of this invention, the step of performing pure main lobe radiation power spectrum analysis based on the high-resolution time-frequency spectrogram to obtain pure main lobe radiation power spectrum information includes: performing noise reduction processing on the high-resolution time-frequency spectrogram to obtain a noise-reduced high-resolution time-frequency spectrogram; extracting the main lobe spectral contour based on the noise-reduced high-resolution time-frequency spectrogram to obtain main lobe spectral contour information; determining the spectral identifier of the canonical frame structure signal based on the main lobe spectral contour information, and performing time-frequency mask analysis based on the spectral identifier to obtain a target time-frequency mask; and performing pure main lobe radiation power spectrum analysis based on the target time-frequency mask to obtain pure main lobe radiation power spectrum information.

[0031] Specifically, a high-resolution time spectrum is determined based on the preprocessed full-band spectrum data. The time-domain signal is mapped to the time-spectrum domain using a Hamming window-weighted short-time Fourier transform based on the preprocessed full-band spectrum data. The window length is set according to one time slot period, and the overlap can be set to 75%, thus obtaining a high-resolution time spectrum containing the complete time evolution.

[0032] Noise reduction processing is performed on the high-resolution time-frequency spectrum. In the high-resolution time-frequency spectrum, a guard band region with no signal transmission is selected. Based on the guard band region, the local minimum envelope of the spectrum is estimated using one-dimensional morphological closing operation. The local minimum envelope is used as the noise basis surface. The high-resolution time-frequency spectrum is subtracted from the noise basis surface point by point to obtain the noise-reduced high-resolution time-frequency spectrum.

[0033] The main lobe spectral contour is extracted based on the high-resolution time-frequency map after noise reduction. The set of continuous frequency points with energy exceeding six dB in the vicinity of the center frequency of each channel is extracted by two-dimensional peak search. Gaussian fitting is performed on each set to obtain the preliminary main lobe spectral contour of each carrier, that is, the main lobe spectral contour information is obtained.

[0034] Based on the main lobe spectral contour information, the spectral identifier of the standard frame structure signal is determined. A two-dimensional cyclic spectral function graph can be determined using the main lobe spectral contour information for cyclic stationary characteristic peak analysis of the carrier signal. Based on the cyclic stationary characteristic peak information, cross-correlation matching is performed using a preset cyclic stationary characteristic peak template library to determine the spectral identifier of the standard frame structure signal using a preset decision threshold. Time-frequency mask analysis is performed based on the spectral identifier. Using the known pure spectral template of the constant envelope sequence in the physical layer frame synchronization header in the frequency domain, a two-dimensional normalized cross-correlation scan is performed on the high-resolution time-frequency spectrum to accurately locate the synchronization time and frequency offset of each frame, obtaining the target time-frequency mask.

[0035] Based on the target time-frequency mask, a pure main lobe radiation power spectrum analysis is performed. The target time-frequency mask is multiplied element-wise with the high-resolution time-frequency spectrum, retaining only the energy distribution within the main lobe frequency band during frame synchronization, and removing all inter-slot leakage and adjacent channel interference components to obtain pure main lobe radiation power spectrum information.

[0036] Furthermore, determining the spectral identifier of the standard frame structure signal based on the main lobe spectral contour information includes: determining a two-dimensional cyclic spectral function graph based on the main lobe spectral contour information, and performing cyclic stationary characteristic peak analysis of the carrier signal based on the two-dimensional cyclic spectral function graph to obtain cyclic stationary characteristic peak information; performing cross-correlation matching based on the cyclic stationary characteristic peak information using a preset cyclic stationary characteristic peak template library to obtain cross-correlation matching results; and determining the spectral identifier of the standard frame structure signal based on the cross-correlation matching results using a preset decision threshold.

[0037] Specifically, a two-dimensional cyclic spectral function graph is determined based on the main lobe spectral contour information. From the preprocessed full-band spectral data, a time-domain baseband signal segment corresponding to the center frequency of the main lobe spectral contour information is extracted. Autocorrelation is performed on the time-domain baseband signal segment, and the autocorrelation function is Fourier transformed along the time axis to obtain the two-dimensional cyclic spectral correlation function graph for each carrier signal. Cyclic stationary characteristic peak analysis of the carrier signal is performed based on the two-dimensional cyclic spectral correlation function graph. A one-dimensional search is conducted along the cyclic frequency axis in the two-dimensional cyclic spectral correlation function graph to extract the spectral correlation peaks appearing at zero cyclic frequency and integer multiples of the symbol rate, thus obtaining the set of cyclic stationary characteristic peak positions and intensities for each carrier signal, i.e., obtaining the cyclic stationary characteristic peak information.

[0038] Based on the cyclic stationary characteristic peak information, cross-correlation matching is performed using a pre-defined cyclic stationary characteristic peak template library. According to the prior knowledge of physical layer standards of cellular communication systems such as Long Term Evolution and New Radio, a cyclic stationary characteristic peak template library unique to the modulation methods of reference signals, synchronization signals, and specific control channels is pre-constructed. The set of cyclic stationary characteristic peak positions and intensities is cross-correlation matched one by one with the templates in the pre-constructed cyclic stationary characteristic peak template library to obtain the template matching similarity of each carrier signal, that is, to obtain the cross-correlation matching result.

[0039] Based on the cross-correlation matching results, the spectral identifier of the standard frame structure signal is determined using a preset decision threshold. A preset decision threshold is set for the template matching similarity of each carrier signal. If the matching similarity between a carrier signal and the template of a cellular communication system exceeds the preset decision threshold, the carrier is determined to be a standard frame structure signal under the cellular communication system, and the corresponding system identifier and frame structure type label are assigned to obtain the spectral identifier of the standard frame structure signal.

[0040] S203: Based on the pure main lobe radiation power spectrum information, perform candidate legitimate base station identity screening to obtain a candidate legitimate base station identity set; In the specific implementation of this invention, candidate legitimate base station identities are screened based on the pure main lobe radiation power spectrum information. Normalized power spectrum envelope morphological features are extracted from the pure main lobe radiation power spectrum information, and dynamic time-normalized distance is calculated with the spectrum fingerprint templates of all legitimate base stations registered in the database for that cell. Several candidate legitimate base station identities with the smallest distance and below the similarity threshold are selected, thus obtaining the candidate legitimate base station identity set.

[0041] S204: Perform posterior probability analysis on each candidate base station based on the candidate legitimate base station identity set to obtain the corresponding posterior probability information; In the specific implementation of this invention, posterior probability analysis of each candidate base station is performed based on the candidate legitimate base station identity set. Combining the received strength at the corresponding location in the flight waypoint position sequence and the pure main lobe radiation power spectrum information, a geographical constraint equation for the difference in arrival signal strength is constructed. Bayesian probability estimation is performed one by one with the base station coordinates in the candidate legitimate base station identity set to calculate the posterior probability that each candidate base station is the true source of the signal, and the corresponding posterior probability information is obtained.

[0042] S205: Perform antenna transmission point attribution analysis based on the posterior probability information to obtain antenna transmission point attribution information; In the specific implementation of this invention, antenna transmission point attribution analysis is performed based on the posterior probability information. If the base station corresponding to the maximum posterior probability is located in the legal registration database and the probability value is higher than a preset threshold, such as 0.9, then the attribution is determined to be normal and the base station identity is recorded; otherwise, it is determined to be an illegal transmission source. Finally, antenna transmission point attribution information containing the estimated coordinates of the suspected fake base station, the signal frequency, and the confidence level of the attribution anomaly can be output.

[0043] S206: Detect antenna transmit power anomalies based on the preprocessed antenna transmit power information and obtain antenna transmit power anomaly detection results; In a specific implementation of this invention, the step of detecting antenna transmit power anomalies based on preprocessed antenna transmit power information to obtain antenna transmit power anomaly detection results includes: performing short-time power mean analysis based on preprocessed antenna transmit power information to obtain a short-time power mean sequence; determining a long-time autocorrelation attenuation coefficient sequence based on preprocessed antenna transmit power information, and determining a target feature vector based on the short-time power mean sequence and the long-time autocorrelation attenuation coefficient sequence; obtaining a reference power curve, and performing antenna transmit power anomaly detection based on the reference power curve and the target feature vector to obtain antenna transmit power anomaly detection results.

[0044] Specifically, short-time power mean analysis is performed based on the preprocessed antenna transmit power information. The power sequence is sliced ​​into ten-second statistical windows from the preprocessed antenna transmit power information, and the arithmetic mean within each window is calculated to obtain the short-time power mean sequence.

[0045] Based on the preprocessed antenna transmit power information, a long-term autocorrelation attenuation coefficient sequence is determined. Autocorrelation analysis is performed on the preprocessed antenna transmit power information to extract the autocorrelation coefficient at a one-hour lag. Combined with the Hearst exponent estimation of the sequence, the long-term autocorrelation attenuation coefficient is calculated. Based on the short-term power mean sequence and the long-term autocorrelation attenuation coefficient sequence, a target feature vector is determined. In the preprocessed antenna transmit power information, the overall standard deviation is calculated for sample points within each ten-second window, and the number of instantaneous power values ​​continuously changing beyond a preset power jump threshold is counted, obtaining a power fluctuation standard deviation sequence and a short-term steep change frequency sequence. The short-term power mean sequence, power fluctuation standard deviation sequence, short-term steep change frequency sequence, and long-term autocorrelation attenuation coefficient sequence are then combined to obtain the target feature vector.

[0046] A baseline power curve is obtained by retrieving historical traffic volume data for the same time period from the operation support system via a network interface. The relationship between traffic volume and historical normal power is fitted to obtain the baseline power curve. Antenna transmit power anomaly detection is performed based on the baseline power curve and the target feature vector. Antenna transmit power offset analysis can be performed using a short-time power mean sequence based on the baseline power curve. Anomaly marking analysis is then performed based on the antenna transmit power offset sequence and the target feature vector to detect antenna transmit power anomalies, thus obtaining the antenna transmit power anomaly detection results.

[0047] Furthermore, the step of detecting antenna transmit power anomalies based on the reference power curve and the target feature vector to obtain antenna transmit power anomaly detection results includes: performing antenna transmit power offset analysis based on the reference power curve using the short-time power mean sequence to obtain an antenna transmit power offset sequence; performing anomaly marking analysis based on the antenna transmit power offset sequence and the target feature vector to obtain anomaly marking sequence; and performing antenna transmit power anomaly detection based on the anomaly marking sequence to obtain antenna transmit power anomaly detection results.

[0048] Specifically, based on the reference power curve, the antenna transmit power offset analysis is performed using the short-time power mean sequence. The short-time power mean sequence is aligned with the reference power curve in time, the difference between the two is calculated point by point, and the difference sequence is subjected to median filtering with a window length of five to filter out isolated occasional deviations, thereby obtaining the antenna transmit power offset sequence.

[0049] Anomaly labeling analysis is performed based on the antenna transmit power offset sequence and target feature vector. The target feature vector and antenna transmit power offset sequence are concatenated into a complete anomaly detection feature set, which is then input into an isolated forest model pre-constructed with a large number of normal operating condition samples. The model calculates the average path length of each sample in the tree through random hyperplane cutting. If the path length is shorter than the anomaly judgment threshold, the time moment is marked as an anomaly, thus obtaining a window-by-window anomaly labeling sequence.

[0050] Antenna transmit power anomaly detection is performed based on the aforementioned anomaly marker sequence. According to the window-by-window anomaly marker sequence combined with the positive and negative directions of power offset, the antenna transmit power anomaly detection result is output, which includes four state types: continuously high power, continuously low power, intermittent drastic jumps, or normal power.

[0051] S207: Analyze the antenna contour region based on the preprocessed base station antenna image, segment the antenna element based on the antenna contour region, and perform attitude recognition based on the antenna element to obtain antenna azimuth angle and downtilt angle attitude information; In the specific implementation of this invention, the step of analyzing the antenna contour region based on the preprocessed base station antenna image, segmenting the antenna elements based on the antenna contour region, and performing attitude recognition based on the antenna elements to obtain antenna azimuth and downtilt attitude information includes: extracting an antenna edge map based on the preprocessed base station antenna image to obtain a target antenna edge map, and determining the antenna contour region based on the target antenna edge map; analyzing a feature map of regular texture differences between antenna element units based on the antenna contour region, and segmenting the antenna elements based on the feature map; obtaining a spatial geometric description of the antenna elements, and performing attitude recognition based on the spatial geometric description to obtain antenna azimuth and downtilt attitude information.

[0052] Specifically, antenna edge map extraction is performed based on the preprocessed base station antenna image. Color space conversion is performed on the preprocessed base station antenna image, and the Laplacian of Gaussian operator is applied to the luminance channel to enhance edges. An adaptive threshold is calculated by fusing the global threshold obtained by Otsu's method with the statistical threshold of the local neighborhood of each pixel, and binarized edge extraction is performed to obtain a fine edge map of the antenna panel boundary, i.e., the target antenna edge map. Based on the target antenna edge map, the antenna contour region is determined. An asymptotic probability Hough transform is run on the target edge map, and random sampling and voting are performed from the edge points to detect all straight line segments whose length exceeds a preset threshold, obtaining a set of candidate straight line segments for the antenna panel. Angle clustering is performed on the candidate straight line segment set, grouping line segments with an angle difference of less than two degrees into one group. The longest vertical and horizontal parallel line pairs in the cluster, whose mutual distance satisfies the panel design size constraints, are identified. Closed polygons are formed by finding the intersections of these line pairs, and the region with the largest inscribed rectangle is taken as the contour region of the antenna panel.

[0053] Based on the antenna contour region, a feature map showing the regular texture differences between antenna elements is analyzed. The antenna elements are then segmented based on this feature map. The preprocessed base station antenna image is cropped into this region according to the antenna contour region and converted into a grayscale image. The local binary mode feature value of each pixel is calculated to obtain a feature map reflecting the regular texture differences between antenna elements on the radome surface. Seed pixels with clear texture in the middle of the antenna are selected from the feature map. Texture consistency and spatial proximity criteria are set, and a region growing algorithm is executed. Pixels that meet the conditions are iteratively absorbed and marked with a mask until the growth boundary stops at the texture abrupt change, thus obtaining an independent antenna element segmentation mask, i.e., the antenna elements are segmented.

[0054] The spatial geometric description of the antenna element is obtained. Based on the antenna element segmentation mask, its contour point set is extracted and fitted with the minimum circumscribed tilted rectangle. At the same time, the Harris corner detection algorithm is used to locate sub-pixel level corners inside the mask to obtain the accurate spatial geometric description of each element. Attitude recognition is performed based on the aforementioned spatial geometric description. The intrinsic parameter matrix and distortion coefficients obtained from camera calibration are called to perform distortion removal and perspective correction on the corner points of the circumscribed rectangle in the spatial geometric description, restoring the true planar orientation of the oscillator in the camera coordinate system. Using the rotation and translation transformation from the camera coordinate system to the UAV body coordinate system and then to the local horizontal coordinate system, the plane normal vector of the oscillator is mapped to the geographic coordinate system to obtain the normal vector direction of the oscillator in the geographic coordinate system. Combined with the yaw angle, gimbal pitch angle and roll angle of the UAV at the moment of shooting, the normal vector direction of the oscillator in the geographic coordinate system is decomposed into the supplementary angle between the horizontal projection direction and the zenith through spatial vector projection. The azimuth angle of the antenna relative to true north and the downtilt angle relative to the horizontal plane are calculated respectively. The results of all oscillator calculations are weighted and fused according to the mask area to obtain the final antenna azimuth angle and downtilt angle attitude information.

[0055] S208: Based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth angle and downtilt angle attitude information, conduct a network security vulnerability survey, obtain network security vulnerability survey results, and generate a survey and inspection report based on the network security vulnerability survey results.

[0056] In the specific implementation of this invention, the step of conducting network security vulnerability investigation based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth angle and downtilt angle attitude information to obtain network security vulnerability investigation results includes: generating a multi-source heterogeneous dataset based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth angle and downtilt angle attitude information; conducting illegal intrusion vulnerability investigation of base station signals based on the multi-source heterogeneous dataset to obtain illegal intrusion vulnerability investigation results; conducting cross-area coverage interference vulnerability investigation based on the multi-source heterogeneous dataset to obtain cross-area coverage interference vulnerability investigation results; and determining the network security vulnerability investigation results based on the illegal intrusion vulnerability investigation results and the cross-area coverage interference vulnerability investigation results.

[0057] Specifically, a multi-source heterogeneous dataset is generated based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth angle and downtilt angle attitude information. The antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth angle and downtilt angle attitude information are mapped onto a unified geographic information system layer according to timestamps and cell identification codes, thereby completing the precise alignment of multi-source heterogeneous data in spatial and temporal dimensions and obtaining an aligned multi-source heterogeneous dataset.

[0058] Based on the aforementioned multi-source heterogeneous dataset, an investigation of illegal intrusion risks of base station signals is conducted. The first type of risk determination is performed on the multi-source heterogeneous dataset. If the antenna transmission point attribution information is marked as an illegal transmission source, an illegal intrusion risk record is directly generated, and an illegal intrusion risk record with a spectrum screenshot and estimated coordinates of the fake base station signal is obtained, thus obtaining the illegal intrusion risk investigation result.

[0059] Based on the aforementioned multi-source heterogeneous dataset, an investigation of cross-area coverage interference risks is conducted. For records with legitimate attribution in the multi-source heterogeneous dataset, a second type of risk assessment is performed. If the antenna transmit power anomaly detection result indicates high power and the antenna azimuth angle points to a sensitive area outside the planned sector coverage angle range, it is determined to be a cross-area coverage interference risk, and a cross-area coverage interference risk record is obtained, thus obtaining the cross-area coverage interference risk investigation result. Based on the illegal intrusion risk investigation result and the cross-area coverage interference risk investigation result, a network security risk investigation result is determined. For records with legitimate attribution in the multi-source heterogeneous dataset, a third type of risk assessment is performed. If the difference between the antenna downtilt angle and the design downtilt angle in the network planning parameters exceeds the allowable threshold, and the antenna transmit power anomaly detection result is accompanied by low power or intermittent jumps, it is determined to be a mechanical downtilt anomaly of the antenna feeder system or a phase shifter fault risk, and a record of the mechanical downtilt anomaly of the antenna feeder system or a phase shifter fault risk is obtained. The results of the illegal intrusion risk investigation, the cross-area coverage interference risk investigation, and the antenna feeder system are then combined. For all potential hazards initially identified in the mechanical tilt anomaly or phase shifter malfunction hazard record, the hazard type, power deviation degree, attitude angle deviation amplitude, illegal attribution confidence level, and duration were selected as evaluation factors. A pairwise comparison matrix was constructed using the analytic hierarchy process (AHP), and the influence weight vector of each factor was calculated. For each hazard, the measured values ​​of each evaluation factor were dimensionless, and the influence weight vectors of each factor were weighted and summed to obtain a comprehensive risk score. The hazards were then classified into three levels—high risk, medium risk, and low risk—according to the score range, resulting in network security hazard investigation results with level labels.

[0060] Based on the network security vulnerability investigation results, an investigation and inspection report is generated. The report uses a pre-set extensible markup language template and inserts a vulnerability summary table with risk level color coding, a comparison chart of pure main lobe radiation power spectrum information, a curve of antenna transmit power offset sequence, a composite image of antenna azimuth and downtilt attitude information, and screenshots of vulnerability locations marked on an electronic map. This automatically fills in a draft investigation and inspection report, which is then converted into a portable document format and pushed with one click via an encrypted transmission channel to a pre-bound maintenance engineer's handheld terminal and the backend network management center. Simultaneously, the report is archived in the inspection history database, resulting in the final archived and pushed investigation and inspection report.

[0061] In this embodiment of the invention, a UAV equipped with a lightweight frequency measurement payload collects full-band spectrum data, antenna transmit power information, and base station antenna images for preprocessing. Based on the preprocessed full-band spectrum data, a clean main lobe radiated power spectrum analysis is performed. Antenna transmission point attribution analysis is then performed based on the clean main lobe radiated power spectrum information to prevent mismatching of side lobes or reflected signals to incorrect antennas, accurately identifying the antenna source of radiated signals. Antenna transmit power anomaly detection is performed based on the preprocessed antenna transmit power information, eliminating the heavy workload of manual power data verification and improving the accuracy of fault identification. Antenna contour regions are analyzed based on the preprocessed base station antenna images, and antenna elements are segmented based on these regions. Attitude recognition is performed based on the antenna elements to obtain antenna azimuth and downtilt attitude information, ensuring the accuracy and consistency of attitude recognition. Network security vulnerability investigation is conducted based on antenna transmission point attribution information, antenna transmit power anomaly detection results, and antenna azimuth and downtilt attitude information. An investigation and inspection report is generated based on the network security vulnerability investigation results, transforming scattered data into operational documents that can directly guide network optimization and maintenance decisions, realizing the value of UAV-based intelligent network investigation and inspection.

[0062] Example 3 Please see Figure 3 , Figure 3 This is a schematic diagram of the structural composition of the UAV intelligent network survey and inspection system in an embodiment of the present invention. The system includes: Preprocessing module 31: Used by the UAV carrying a lightweight frequency measurement payload to collect full-band spectrum data, antenna transmit power information and base station antenna image, and to preprocess the full-band spectrum data, antenna transmit power information and base station antenna image to obtain preprocessed full-band spectrum data, antenna transmit power information and base station antenna image; Transmission point attribution module 32: is used to perform pure main lobe radiated power spectrum analysis based on preprocessed full-band spectrum data to obtain pure main lobe radiated power spectrum information, and to perform antenna transmission point attribution analysis based on the pure main lobe radiated power spectrum information to obtain antenna transmission point attribution information. Anomaly detection module 33: used to perform antenna transmit power anomaly detection based on preprocessed antenna transmit power information, and obtain antenna transmit power anomaly detection results; Attitude recognition module 34: used to analyze the antenna contour region based on the preprocessed base station antenna image, segment the antenna element based on the antenna contour region, and perform attitude recognition based on the antenna element to obtain antenna orientation angle and downtilt angle attitude information; Report generation module 35: is used to conduct network security vulnerability investigation based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth angle and downtilt angle attitude information, obtain network security vulnerability investigation results, and generate an investigation and inspection report based on the network security vulnerability investigation results.

[0063] In the specific implementation of this invention, the specific implementation methods of the system items can be referred to the implementation methods of the above-mentioned method items, and will not be repeated here.

[0064] In this embodiment of the invention, a UAV equipped with a lightweight frequency measurement payload collects full-band spectrum data, antenna transmit power information, and base station antenna images for preprocessing. Based on the preprocessed full-band spectrum data, a clean main lobe radiated power spectrum analysis is performed. Antenna transmission point attribution analysis is then performed based on the clean main lobe radiated power spectrum information to prevent mismatching of side lobes or reflected signals to incorrect antennas, accurately identifying the antenna source of radiated signals. Antenna transmit power anomaly detection is performed based on the preprocessed antenna transmit power information, eliminating the heavy workload of manual power data verification and improving the accuracy of fault identification. Antenna contour regions are analyzed based on the preprocessed base station antenna images, and antenna elements are segmented based on these regions. Attitude recognition is performed based on the antenna elements to obtain antenna azimuth and downtilt attitude information, ensuring the accuracy and consistency of attitude recognition. Network security vulnerability investigation is conducted based on antenna transmission point attribution information, antenna transmit power anomaly detection results, and antenna azimuth and downtilt attitude information. An investigation and inspection report is generated based on the network security vulnerability investigation results, transforming scattered data into operational documents that can directly guide network optimization and maintenance decisions, realizing the value of UAV-based intelligent network investigation and inspection.

[0065] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0066] Furthermore, the above provides a detailed description of the UAV intelligent network survey and inspection method and system provided by the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for intelligent network surveying and inspection using unmanned aerial vehicles (UAVs), characterized in that, The method includes: The UAV carries a lightweight frequency measurement payload to collect full-band spectrum data, antenna transmit power information, and base station antenna images. It also preprocesses the full-band spectrum data, antenna transmit power information, and base station antenna images to obtain preprocessed full-band spectrum data, antenna transmit power information, and base station antenna images. Based on the preprocessed full-band spectrum data, a pure main lobe radiated power spectrum analysis is performed to obtain pure main lobe radiated power spectrum information. Based on the pure main lobe radiated power spectrum information, an antenna transmission point attribution analysis is performed to obtain antenna transmission point attribution information. Antenna transmit power anomaly detection is performed based on the preprocessed antenna transmit power information to obtain antenna transmit power anomaly detection results; The antenna contour region is analyzed based on the preprocessed base station antenna image, and the antenna elements are segmented based on the antenna contour region. Attitude recognition is performed based on the antenna elements to obtain the antenna azimuth angle and downtilt angle attitude information. Based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth angle and downtilt angle attitude information, a network security vulnerability investigation is conducted to obtain network security vulnerability investigation results, and an investigation and inspection report is generated based on the network security vulnerability investigation results.

2. The UAV intelligent network survey and inspection method according to claim 1, characterized in that, The preprocessing of the full-band spectrum data, antenna transmit power information, and base station antenna image to obtain preprocessed full-band spectrum data, antenna transmit power information, and base station antenna image includes: Harmonic suppression is performed on the full-band spectrum data to obtain harmonic-suppressed full-band spectrum data, and multi-scale wavelet packet decomposition is performed on the harmonic-suppressed full-band spectrum data to obtain preprocessed full-band spectrum data. The antenna transmit power information is cleaned to obtain preprocessed antenna transmit power information; The base station antenna image is dehazed to obtain a dehazed base station antenna image, and then geometric correction is performed on the dehazed base station antenna image to obtain a preprocessed base station antenna image.

3. The UAV intelligent network survey and inspection method according to claim 1, characterized in that, The process involves performing a clean main lobe radiated power spectrum analysis based on the preprocessed full-band spectrum data to obtain clean main lobe radiated power spectrum information, and then performing an antenna transmission point attribution analysis based on the clean main lobe radiated power spectrum information to obtain antenna transmission point attribution information, including: A high-resolution time-frequency spectrum is determined based on the preprocessed full-band spectrum data, and a clean main lobe radiation power spectrum analysis is performed based on the high-resolution time-frequency spectrum to obtain clean main lobe radiation power spectrum information. Based on the pure main lobe radiation power spectrum information, candidate legitimate base station identities are screened to obtain a set of candidate legitimate base station identities. Based on the set of candidate legitimate base station identities, perform posterior probability analysis on each candidate base station to obtain the corresponding posterior probability information; Antenna transmission point attribution analysis is performed based on the posterior probability information to obtain antenna transmission point attribution information.

4. The UAV intelligent network survey and inspection method according to claim 3, characterized in that, The step of performing a clean main lobe radiation power spectrum analysis based on the high-resolution time-frequency spectrogram to obtain clean main lobe radiation power spectrum information includes: The high-resolution time-frequency spectrum is denoised to obtain the denoised high-resolution time-frequency spectrum. The main lobe spectral contour is extracted based on the high-resolution time-spectral map after noise reduction to obtain the main lobe spectral contour information. Based on the main lobe spectral contour information, the spectral identifier of the standard frame structure signal is determined, and time-frequency mask analysis is performed based on the spectral identifier to obtain the target time-frequency mask; Based on the target time-frequency mask, a pure main lobe radiation power spectrum analysis is performed to obtain pure main lobe radiation power spectrum information.

5. The UAV intelligent network survey and inspection method according to claim 4, characterized in that, The step of determining the spectral identifier of the canonical frame structure signal based on the main lobe spectral contour information includes: Based on the main lobe spectral contour information, a two-dimensional cyclic spectral function graph is determined, and based on the two-dimensional cyclic spectral function graph, the cyclic stationary characteristic peak analysis of the carrier signal is performed to obtain cyclic stationary characteristic peak information. Based on the cyclic stationary characteristic peak information, cross-correlation matching is performed using a preset cyclic stationary characteristic peak template library to obtain cross-correlation matching results. Based on the cross-correlation matching results, the spectral identifier of the standard frame structure signal is determined using a preset decision threshold.

6. The UAV intelligent network survey and inspection method according to claim 1, characterized in that, The antenna transmit power anomaly detection based on the preprocessed antenna transmit power information, to obtain the antenna transmit power anomaly detection result, includes: Short-time power mean analysis is performed based on the preprocessed antenna transmit power information to obtain a short-time power mean sequence. The long-term autocorrelation attenuation coefficient sequence is determined based on the preprocessed antenna transmit power information, and the target feature vector is determined based on the short-term power mean sequence and the long-term autocorrelation attenuation coefficient sequence. A reference power curve is obtained, and antenna transmit power anomaly detection is performed based on the reference power curve and the target feature vector to obtain the antenna transmit power anomaly detection result.

7. The UAV intelligent network survey and inspection method according to claim 6, characterized in that, The process of detecting antenna transmit power anomalies based on the reference power curve and the target feature vector, and obtaining antenna transmit power anomaly detection results, includes: Based on the reference power curve, antenna transmit power offset analysis is performed using the short-time power mean sequence to obtain the antenna transmit power offset sequence. Anomaly marking analysis is performed based on the antenna transmit power offset sequence and target feature vector to obtain anomaly marking sequence; Antenna transmit power anomaly detection is performed based on the anomaly marker sequence to obtain antenna transmit power anomaly detection results.

8. The UAV intelligent network survey and inspection method according to claim 1, characterized in that, The process of analyzing the antenna contour region based on the preprocessed base station antenna image, segmenting the antenna elements based on the antenna contour region, and performing attitude recognition based on the antenna elements to obtain antenna azimuth and downtilt attitude information includes: Antenna edge map extraction is performed based on the preprocessed base station antenna image to obtain the target antenna edge map, and the antenna contour region is determined based on the target antenna edge map; The antenna element features are analyzed based on the antenna contour region to identify the regular texture differences between the antenna elements, and the antenna elements are segmented based on the feature maps. The spatial geometric description of the antenna element is obtained, and attitude recognition is performed based on the spatial geometric description to obtain the antenna orientation angle and downtilt angle attitude information.

9. The UAV intelligent network survey and inspection method according to claim 1, characterized in that, The network security vulnerability investigation is conducted based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth and downtilt attitude information to obtain network security vulnerability investigation results, including: A multi-source heterogeneous dataset is generated based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, and antenna azimuth and downtilt attitude information. Based on the aforementioned multi-source heterogeneous dataset, an investigation of potential illegal intrusion risks into base station signals is conducted, and the results of the investigation are obtained. Based on the multi-source heterogeneous dataset, a cross-area coverage interference risk investigation is conducted to obtain the cross-area coverage interference risk investigation results. Based on the illegal intrusion risk investigation results and the cross-area coverage interference risk investigation results, the network security risk investigation results are determined.

10. A UAV intelligent network survey and inspection system, characterized in that, The system includes: Preprocessing module: Used by UAV carrying a lightweight frequency measurement payload to collect full-band spectrum data, antenna transmit power information and base station antenna images, and to preprocess the full-band spectrum data, antenna transmit power information and base station antenna images to obtain preprocessed full-band spectrum data, antenna transmit power information and base station antenna images; Transmitter point attribution module: used to perform pure main lobe radiated power spectrum analysis based on preprocessed full-band spectrum data to obtain pure main lobe radiated power spectrum information, and to perform antenna transmitter point attribution analysis based on the pure main lobe radiated power spectrum information to obtain antenna transmitter point attribution information; Anomaly detection module: used to detect antenna transmit power anomalies based on preprocessed antenna transmit power information and obtain antenna transmit power anomaly detection results; Attitude recognition module: used to analyze the antenna contour region based on the preprocessed base station antenna image, segment the antenna element based on the antenna contour region, and perform attitude recognition based on the antenna element to obtain antenna azimuth angle and downtilt angle attitude information; Report generation module: used to conduct network security vulnerability investigation based on the antenna transmission point attribution information, antenna transmission power anomaly detection results, antenna azimuth angle and downtilt angle attitude information, obtain network security vulnerability investigation results, and generate an investigation and inspection report based on the network security vulnerability investigation results.

Images