Radio frequency signal source positioning and geographic coordinate mapping method based on single unmanned aerial vehicle platform

Abstract

The application discloses a radio frequency signal source positioning and geographic coordinate mapping method based on a single unmanned aerial vehicle platform, and relates to the technical field of radio frequency positioning.The radio frequency signal source positioning and geographic coordinate mapping method based on the single unmanned aerial vehicle platform comprises the following steps: S1, collecting unmanned aerial vehicle maneuvering observation data and performing pretreatment; S2, constructing a virtual multi-base station candidate observation set; S3, screening a selected virtual base station point set, and identifying an observation geometric degeneration state and a degeneration reinforcement direction; S4, constructing three types of observation constraints; generating a target position estimation value, a non-line-of-sight high-risk sector and a next optimal observation navigation point; and S5, determining whether to terminate the maneuvering radio frequency observation, and outputting a WGS84 geographic coordinate.The method solves the problem that the existing single unmanned aerial vehicle radio frequency positioning method is difficult to distinguish the observation quality and geometric contribution of multi-time slot observation points, causes the same side repeated observation to continuously participate in joint positioning, and further causes the problems of insufficient convergence of the target position uncertainty and poor stability of the geographic coordinate output.

Description

Technical Field

[0001] This invention relates to the field of radio frequency positioning technology, specifically to a method for radio frequency signal source positioning and geographic coordinate mapping based on a single unmanned aerial vehicle (UAV) platform. Background Technology

[0002] With the widespread application of wireless communication equipment, emergency beacons, IoT terminals, and low-power radio frequency (RF) transmitters in scenarios such as field patrols, emergency rescue, asset monitoring, and ecological monitoring, RF signal source localization technology has gradually become an important technical means for target search, abnormal signal investigation, and spatial situational awareness. Traditional RF localization typically relies on fixed receiving stations, ground direction-finding equipment, or multi-node collaborative receiving networks. It estimates the signal source location by analyzing signal strength differences, time differences of arrival, angles of arrival, or frequency shifts between multiple receiving locations, and then uses a geographic information system (GIS) to complete spatial coordinate labeling. However, in mountainous forest areas, wilderness areas, disaster sites, and temporary mission areas, fixed ground receiving stations are often difficult to deploy in a timely manner, and ground communication links and power supply conditions are easily limited by terrain, environment, and accessibility. In recent years, UAV platforms equipped with RF receiving equipment have been gradually used for RF signal search and localization tasks in conditions without fixed infrastructure due to their high mobility, rapid deployment, and flexible coverage. By conducting multiple observations from different spatial locations using UAVs, RF measurement information from multiple perspectives can be obtained, providing a data foundation for ground RF signal source localization and geographic coordinate mapping. Therefore, how to use UAV maneuver observation results to achieve stable and reliable radio frequency signal source localization under complex field conditions has become an important research direction in the field of UAV remote sensing monitoring and radio frequency positioning.

[0003] For example, the invention patent with announcement number CN107390166B discloses an adaptive interference source localization flight verification method, which can monitor and locate various ground-based radio interference sources to ensure civil aviation radio safety. This method uses airborne radio detection equipment to obtain the azimuth of the interference source and combines this with the aircraft's own flight parameters such as position, speed, flight attitude, and altitude. It then applies the single-station cross-positioning principle and adaptive filtering method to obtain the interference source location information. The main steps include: establishing the state equation of the interference detection system of the verification aircraft; establishing the nonlinear measurement equation of the interference detection system of the verification aircraft; discretizing the system state equation and measurement equation; and using an adaptive interference source localization algorithm to solve the discretized equations to output the location of the interference source. This method improves the localization accuracy and stability of the localization process through the adaptive interference source localization algorithm.

[0004] For example, invention patent CN108832994B discloses a helicopter-based radio monitoring system, including an airborne device and a ground monitoring device. The airborne device includes a helicopter platform and radio monitoring equipment for monitoring radio signals in the air. The ground monitoring device is connected to the airborne device and controls the airborne device based on its monitoring results. This system, based on a helicopter platform carrying radio monitoring equipment, performs airborne radio monitoring tasks, enabling rapid and accurate direction finding and location of radio transmission signals, thus reducing the workload of radio monitoring personnel.

[0005] Existing airborne platform radio frequency (RF) signal source localization technologies can acquire RF observation information using airborne radio detection or monitoring equipment, and combine this information with parameters such as the flight platform's position, speed, and attitude to determine the direction and location of ground-based radio transmitters or interference sources. However, these methods typically focus more on acquiring observation data, single-station cross-location, filtering, or deploying airborne monitoring systems, with insufficient attention paid to whether candidate observation points formed at multiple measurement times on a single platform truly improve the positioning geometry. In scenarios without fixed ground receiving stations, with terrain obstruction, and limited flight paths, multiple observation points tend to concentrate in areas on the same side or in similar headings. Even if the observation signal quality is good, it may contribute little to the convergence of target position uncertainty, causing the positioning result to fluctuate along a specific direction after an increase in the number of measurements. This, in turn, affects the stability of the RF signal source's geographic coordinate output and the efficiency of field remeasurement.

[0006] Therefore, in order to address the above problems, there is an urgent need for a method for radio frequency signal source localization and geographic coordinate mapping based on a single UAV platform. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a method for radio frequency signal source localization and geographic coordinate mapping based on a single UAV platform. This method solves the problem that existing single UAV radio frequency localization methods are unable to distinguish the observation quality and geometric contribution of multiple time slot observation points, leading to repeated observations on the same side continuously participating in joint localization, which in turn causes insufficient convergence of target position uncertainty and poor stability of geographic coordinate output.

[0009] Technical solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: a method for RF signal source localization and geographic coordinate mapping based on a single UAV platform, comprising: S1, collecting and preprocessing UAV maneuvering observation data during the single UAV's maneuvering flight; S2, constructing a virtual multi-base station candidate observation set based on the UAV maneuvering observation data; filtering candidate virtual base stations based on RF fingerprint, Doppler, and track matching results, as well as an observation quality score generated by time synchronization deviation values, to obtain a homogeneous candidate observation set; S3, constructing a contribution evaluation number based on the homogeneous candidate observation set. The data collection process involves: S4, selecting virtual base stations based on contribution coefficients and identifying the geometric degradation state and reinforcement direction of observations; S5, constructing three types of observation constraints based on the selected virtual base station set; generating target position estimates, posterior covariance matrices, non-line-of-sight high-risk sectors, and the next optimal waypoint through joint positioning solutions; S6, determining whether to terminate maneuver radio frequency observations by judging the diminishing relationship between the positioning posterior variance, geometric accuracy factor, normalized residual mean square, and new observation sampling gains, and outputting WGS84 geographic coordinates, altitude, and positioning accuracy information after the acquisition is terminated.

[0011] Furthermore, the specific process of collecting UAV maneuver observation data during single UAV maneuver flight is as follows: A single UAV platform is equipped with a wideband radio frequency receiver, an RTK-GNSS positioning module, an IMU inertial measurement unit, and an airborne edge computing unit. At multiple measurement moments, the single UAV is controlled to reach different spatial positions along a planar flight trajectory. During the single UAV maneuver flight, UAV maneuver observation data is collected. The UAV maneuver observation data includes: the received signal strength, time difference of arrival, Doppler frequency shift, radio frequency fingerprint features, and radio frequency receiver sampling timestamps collected by the wideband radio frequency receiver; the platform's three-dimensional position coordinates, platform velocity vector, positioning module timing timestamp, and platform positioning accuracy values ​​collected by the RTK-GNSS positioning module; and the platform's attitude angles collected by the IMU inertial measurement unit.

[0012] Further, the specific preprocessing process is as follows: the platform's three-dimensional position coordinates, platform velocity vector, platform attitude angle, received signal strength, time difference of arrival, Doppler frequency shift, radio frequency fingerprint features, radio frequency receiver sampling timestamp, and positioning module timing timestamp at the same measurement time are time-bound; data records with missing timestamps, missing platform three-dimensional position coordinates, and missing radio frequency measurement fields are removed; the platform's three-dimensional position coordinates are transformed to a unified local coordinate system; and the received signal strength, time difference of arrival, Doppler frequency shift, platform velocity vector magnitude, and platform attitude angle are Z-score standardized.

[0013] Furthermore, the specific process of constructing a virtual multi-base station candidate observation set based on UAV mobile observation data is as follows: the platform's three-dimensional position coordinates are paired with the received signal strength, time difference of arrival, Doppler frequency shift, and radio frequency fingerprint features according to the measurement time. The platform's three-dimensional position coordinates at each measurement time are marked as a candidate virtual base station. The candidate virtual base station and the received signal strength, time difference of arrival, Doppler frequency shift, and radio frequency fingerprint features at the same measurement time are written into the virtual multi-base station candidate observation set.

[0014] Furthermore, based on the RF fingerprint, Doppler, and track matching results, and using the observation quality score generated by the time synchronization deviation value to screen candidate virtual base stations, the specific process for obtaining the homogeneous candidate observation set is as follows: The RF fingerprint features of adjacent measurement times are matched for spectral peak positions, bandwidth profiles, and energy distribution similarity to obtain the RF fingerprint matching results; the Doppler frequency shift, platform velocity vector, and platform three-dimensional position coordinates of adjacent measurement times are checked for residuals to obtain the Doppler matching results; the platform three-dimensional position coordinates, platform velocity vector, and platform attitude angles of the previous measurement time are extrapolated to the current measurement time, and position deviation is checked against the platform three-dimensional position coordinates of the current measurement time to obtain the track matching results; the RF fingerprint features of adjacent measurement times are matched for spectral peak positions, bandwidth profiles, and energy distribution similarity to obtain the RF fingerprint matching results; the platform velocity vector, platform attitude angles of the previous measurement time are extrapolated to the current measurement time, and position deviation is checked against the platform three-dimensional position coordinates of the current measurement time to obtain the track matching results; the platform velocity vector, platform attitude angles of the previous measurement time are extrapolated to the current measurement time, and the platform attitude angles of the previous measurement time are checked for position deviation to obtain the track matching results; the platform velocity vector, platform attitude angles, platform attitude angles, and platform attitude angles are checked for residuals to obtain the RF fingerprint ... platform attitude angles, platform attitude angles, platform attitude angles, platform attitude angles, platform attitude angles, platform attitude angles, platform attitude angles, platform attitude angles, The absolute value of the difference between the sampling timestamp of the frequency receiver and the timing timestamp of the positioning module is used as the time synchronization deviation value. The inter-frame jump value of the received signal strength, the residual dispersion of the time difference of arrival, the residual dispersion of the Doppler frequency shift, and the time synchronization deviation value are normalized and fused to obtain the observation quality score. The observation quality score characterizes the signal stability and time binding reliability of a single measurement record. The radio frequency fingerprint matching results, Doppler matching results, track matching results, and observation quality scores of candidate virtual base stations are used to determine outliers based on interquartile range. Candidate virtual base stations that do not constitute outliers are retained in the same source candidate observation set, while candidate virtual base stations that constitute outliers are excluded from the data range of the current joint positioning solution.

[0015] Furthermore, the specific process of constructing a contribution evaluation dataset by combining the candidate observation set from the same source is as follows: Based on the measurement time, the first q candidate virtual base stations are selected from the candidate observation set from the same source. The distance between the platform's three-dimensional position coordinates of any two selected candidate virtual base stations exceeds the larger of the platform positioning accuracy values ​​at the two measurement times, and these are added to the selected virtual base station set. Before subsequent candidate virtual base stations enter the contribution judgment, the candidate virtual base stations are merged with the selected virtual base station set to form the observation dataset to be updated. Based on the baseline direction between the platform's three-dimensional position coordinates in the candidate virtual base station and the selected virtual base station set, the change in the minimum included angle of the spatial distribution before and after the addition of the candidate virtual base station is calculated, obtaining the geometric angle change. The geometric angle change is used to represent... The degree to which the addition of candidate virtual base stations changes the spatial angle of the selected virtual base station set is determined. Based on the platform 3D position coordinate distance, baseline direction angle, and RF fingerprint feature similarity between candidate and selected virtual base stations, the distance proximity relationship, direction repetition relationship, and signal feature repetition relationship are determined. These relationships are then normalized and fused according to the degree of same-direction aggregation to generate a redundancy level, which is used to characterize the degree of spatial and signal feature repetition between candidate and selected virtual base stations. The observation dataset to be updated is used as the contribution evaluation dataset. The Fisher information matrix is ​​updated based on the observation equation formed by the received signal strength, time difference of arrival, and Doppler frequency shift. The decrease ratio of the posterior covariance matrix trace before and after the combination is calculated to obtain the posterior covariance decrease.

[0016] Further, the specific process of selecting the virtual base station set through contribution coefficients and identifying the observed geometric degradation state and degradation reinforcement direction is as follows: The change in geometric angle, redundancy, and decrease in posterior covariance are input into the contribution discrimination function to generate contribution coefficients. The contribution discrimination function is a function that determines the contribution coefficients of candidate virtual base stations based on the change in geometric angle, redundancy, and decrease in posterior covariance. The contribution discrimination function uses the product term of the change in geometric angle and the decrease in posterior covariance to represent the gain contribution, and uses an exponential decay term of redundancy to suppress repeated observation contributions. When the contribution coefficient reaches the contribution threshold, the candidate virtual base station is added to the selected virtual base station set; when the contribution coefficient does not reach the contribution threshold, the candidate virtual base station is added to the redundant candidate observation record. A spatial distribution matrix is ​​constructed based on the three-dimensional position coordinates of the platform in the selected virtual base station set, and principal component analysis is used to obtain the eigenvalues ​​of the first direction and the second direction. The eigenvalues ​​and the main distribution direction of the base station are used for eigenvalue decomposition. The posterior covariance matrix obtained by inverting the Fisher information matrix during the first observation geometry degradation identification, or the previous round's posterior covariance matrix read during subsequent observation geometry degradation identification, is used as the direction of the major axis of the posterior covariance ellipse. When the ratio of the eigenvalues ​​of the first direction to the eigenvalues ​​of the second direction reaches the geometry degradation judgment threshold, and the angle between the main distribution direction of the base station and the major axis of the posterior covariance ellipse is not greater than the degradation angle threshold, the current observation geometry is determined to be in a degraded state. Among the candidate normal directions on both sides of the major axis of the posterior covariance ellipse, the direction that maximizes the decrease in the predicted posterior covariance matrix trace is selected as the degradation reinforcement direction, constraining the next optimal observation waypoint to move towards a spatial orientation that reduces the uncertainty of the major axis of the posterior covariance ellipse. Otherwise, the current observation geometry is determined to be in a non-degraded state, and the degradation reinforcement direction is set to empty.

[0017] Furthermore, the specific process of constructing three types of observation constraints based on the selected set of virtual base stations is as follows: Based on the selected set of virtual base stations, three types of observation constraints are constructed: the received signal strength and the platform's three-dimensional position coordinates are converted into a received signal strength ranging constraint according to the free space path loss relationship; the arrival time difference and the platform's three-dimensional position coordinates are converted into an arrival time difference hyperbolic constraint according to the distance difference relationship; the Doppler frequency shift, the platform velocity vector, and the platform's three-dimensional position coordinates are converted into a Doppler frequency shift azimuth constraint according to the velocity projection relationship; based on the time synchronization deviation value, the change in the platform velocity vector direction, and the outlier results of the observation residuals of the three types of observation constraints obtained from the previous round of joint positioning calculation or the current round of pre-calculation, the three types of observation constraints are respectively set to an enabled state, a reduced-weight state, or a suspended state, as the participation states of the three types of observation constraints.

[0018] Furthermore, the specific process of generating the target position estimate, posterior covariance matrix, non-line-of-sight high-risk sector, and next optimal observation waypoint through joint positioning solution is as follows: Observation constraints entering the joint positioning solution are selected according to their participation status among the three types of observation constraints, and the fusion weights are determined based on the traces of the posterior covariance matrices of the three types of observation constraints to obtain the target position estimate and posterior covariance matrix; non-line-of-sight observations are identified based on the observation residuals obtained from the joint positioning solution, and the observation residuals of the received signal strength ranging constraint and the time difference hyperbolic constraint in n consecutive measurement times are used as virtual base stations for positive outlier observations. These virtual base stations are connected to the target position estimate, the anomalous observation azimuth is calculated, and the result is generated based on the coverage area of ​​the anomalous observation azimuth. Non-line-of-sight (NLS) high-risk sectors are established to record the spatial azimuth range of continuous positive outlier observations. A set of candidate waypoints is generated based on the major axis direction of the posterior covariance ellipse, the degradation state of the observation geometry, the degradation reinforcement direction, and the NLS high-risk sectors. Candidate waypoints located within the NLS high-risk sectors are excluded, and when the current observation geometry is in a degradation state, they are sorted in ascending order of the angle with the degradation reinforcement direction. For the sorted candidate waypoints, the decrease in the posterior covariance matrix trace caused by the formation of new candidate virtual base stations and their addition to the selected virtual base station set is calculated, along with the turning angle and range increment from the current waypoint to the candidate waypoint. Finally, the next optimal waypoint and the next flight path are determined.

[0019] Furthermore, the specific process of determining whether to terminate maneuvering radio frequency observations and outputting WGS84 geographic coordinates, altitude, and positioning accuracy information after the acquisition is terminated is as follows: Based on the posterior covariance matrix, the platform's three-dimensional position coordinates in the selected virtual base station set, the observation residuals, and the change in the posterior covariance matrix trace before and after adding the next optimal observation waypoint to the selected virtual base station set, the posterior covariance matrix trace of the selected virtual base station set is obtained, respectively. Sampling gains; if the posterior variance, geometric precision factor, and normalized residual mean square do not satisfy the decreasing relationship within m consecutive positioning calculation rounds, or if the gains from new observation sampling are not higher than the gains from new observation sampling in the previous round for m consecutive rounds, then the maneuvering radio frequency observation is terminated; otherwise, data acquisition continues at the next optimal waypoint; after data acquisition is terminated, the target position estimate is converted from the local coordinate system to the WGS84 geodetic coordinate system, and the WGS84 geographic coordinates, altitude, and positioning accuracy information of the ground radio frequency signal source are output. The positioning accuracy information includes circular probability error (CEP), root mean square positioning error (RMS), and target position covariance ellipse parameters.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) This invention collects UAV maneuver observation data during the maneuver flight of a single UAV, and performs joint positioning calculation and WGS84 geographic coordinate mapping based on the selected set of virtual base stations. This enables the radio frequency signal source positioning process to complete spatial position estimation and geographic coordinate output under the condition of no fixed ground receiving station, thereby realizing the effect of autonomous positioning and coordinate mapping of the single UAV radio frequency signal source. This effectively solves the problem of limited complex field deployment caused by reliance on fixed receiving stations in the prior art.

[0023] (2) This invention constructs a virtual multi-base station candidate observation set and filters candidate virtual base stations based on radio frequency fingerprint matching results, Doppler matching results, track matching results and observation quality scores to obtain a homogeneous candidate observation set, thereby achieving the effect of homogeneous screening of multi-time slot observation data and removal of abnormal observations, effectively solving the problem that outlier observations are easily mixed into observation data at different measurement times in the prior art.

[0024] (3) In this invention, contribution coefficients are generated by the change in geometric angle, redundancy and decrease in posterior covariance, and the set of virtual base stations is selected based on the contribution coefficients. This achieves the effect of screening candidate virtual base stations from observation quality to positioning contribution, effectively solving the problem of insufficient convergence of target position uncertainty in the existing technology due to repeated observations on the same side continuously participating in joint positioning.

[0025] (4) The present invention identifies the observed geometric degradation state and determines the degradation reinforcement direction among the candidate normal directions on both sides of the major axis of the posterior covariance ellipse, thereby achieving the directional reinforcement effect of the observed geometric degradation direction, effectively solving the problem of the location result drifting along a specific direction due to the concentrated spatial distribution of virtual base stations in the prior art.

[0026] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0027] Figure 1 This is a flowchart of the radio frequency signal source localization and geographic coordinate mapping method based on a single UAV platform according to the present invention;

[0028] Figure 2 This is a flowchart of the observation geometry degradation identification and degradation reinforcement direction generation process of the present invention;

[0029] Figure 3 This is a schematic diagram of the observation geometry degradation identification and reverse waypoint reinforcement of the present invention;

[0030] Figure 4 This is a comparison chart showing the engineering verification results of the present invention. Detailed Implementation

[0031] 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.

[0032] Please see Figures 1-3 This invention provides a technical solution: a method for radio frequency signal source localization and geographic coordinate mapping based on a single UAV platform, comprising the following steps: S1, collecting UAV maneuvering observation data and preprocessing it during the maneuvering flight of a single UAV; S2, constructing a virtual multi-base station candidate observation set based on the UAV maneuvering observation data; filtering candidate virtual base stations based on radio frequency fingerprint, Doppler and track matching results, and observation quality scores generated by time synchronization deviation values, to obtain a homogeneous candidate observation set; S3, constructing a contribution evaluation dataset by combining the homogeneous candidate observation set; filtering the selected virtual base station set by contribution coefficients, and identifying the observation geometric degradation state and degradation reinforcement direction; S4, constructing three types of observation constraints based on the selected virtual base station set; generating target position estimates, posterior covariance matrices, non-line-of-sight high-risk sectors and the next optimal observation waypoint through joint positioning solutions; S5, determining whether to terminate maneuvering radio frequency observation by judging the decreasing relationship between the positioning posterior variance, geometric accuracy factor, normalized residual mean square, and new observation sampling gains, and outputting WGS84 geographic coordinates, altitude, and positioning accuracy information after the acquisition is terminated.

[0033] Specifically, the process of collecting UAV maneuvering observation data during single UAV maneuvering flight is as follows: A single UAV platform is equipped with a wideband RF receiver, an RTK-GNSS positioning module, an IMU inertial measurement unit, and an airborne edge computing unit. This method is executed locally on the single UAV platform by the airborne edge computing unit. At multiple measurement moments, the single UAV is controlled to reach different spatial positions along a planar flight trajectory. The planar flight trajectory is generated based on the mission area boundary, no-fly zone constraints, minimum safe flight altitude, and maximum flight speed, ensuring that multiple measurement moments correspond to different platform spatial positions. During the single UAV maneuvering flight, UAV maneuvering observation data is collected. This data includes: received signal strength, time difference of arrival, Doppler frequency shift, RF fingerprint characteristics, and RF receiver sampling timestamps collected by a wideband RF receiver operating in a frequency band covering the ground-based RF signal source. The received signal strength is output from the received power measurement channel of the wideband RF receiver. The time difference of arrival is calculated by comparing the arrival times of the characteristic peaks of the same source RF signal at different measurement moments with the reference measurement time. The arrival time difference is obtained, and the arrival times of the characteristic peaks at different measurement times are aligned with the time reference provided by the positioning module's time stamp. The Doppler frequency shift is obtained from the carrier frequency offset tracking results, and the radio frequency fingerprint features are extracted from the spectral peak position, bandwidth profile, and energy distribution. The platform's three-dimensional position coordinates, platform velocity vector, positioning module time stamp, and platform positioning accuracy value are acquired through the RTK-GNSS positioning module. The RTK-GNSS positioning module is the real-time dynamic carrier phase differential global navigation satellite system positioning module. The platform's three-dimensional position coordinates and platform velocity vector are output by the RTK-GNSS positioning module at the same measurement time and are used to calibrate the spatial position and motion state of candidate virtual base stations. The platform's attitude angles are acquired through the IMU inertial measurement unit. The platform attitude angles include the UAV's pitch angle, yaw angle, and roll angle at the measurement time and are used to correct the influence of the platform's motion attitude on the radio frequency observation direction. The platform positioning accuracy value is output by the RTK-GNSS positioning module along with the platform's three-dimensional position coordinates and is used to characterize the platform position calculation accuracy at the corresponding measurement time.

[0034] In this embodiment, a single UAV is a hexacopter industrial UAV, which is equipped with a wideband RF receiver operating in the 100MHz to 6GHz frequency band, an RTK-GNSS positioning module, a nine-axis IMU inertial measurement unit, and an airborne edge computing unit. The airborne edge computing unit includes a field-programmable gate array (FPGA) processing unit and an embedded ARM processing unit. The FPGA processing unit is used to complete RF sampling buffering, RF receiver sampling timestamp attachment, and preliminary RF fingerprint feature extraction. The embedded ARM processing unit is used to perform preprocessing, virtual multi-base station candidate observation set construction, homogeneous candidate observation set screening, contribution coefficient calculation, observation geometric degradation state identification, three types of observation constraint construction, joint positioning solution, next optimal observation waypoint generation, and WGS84 geographic coordinate output. S1 to S5 are executed on the UAV's local edge computing, completing the positioning solution without relying on background offline simulation. The ground terminal is used to receive WGS84 geographic coordinates, altitude, and positioning accuracy information.

[0035] This implementation plan establishes a UAV maneuver observation data foundation with clear spatiotemporal calibration, complete observation fields, and traceable motion status. This ensures that each measurement moment corresponds to a determined platform spatial location, platform motion status, and radio frequency observation results. It provides a unified data source for the spatial positioning of candidate virtual base stations, radio frequency feature acquisition, subsequent homogeneous source screening, observation quality evaluation, and joint positioning calculation. This effectively avoids deviations in virtual base station construction caused by unclear position calibration, inconsistent sampling times, or missing observation fields during multi-timeslot observations by a single UAV.

[0036] Specifically, the preprocessing process involves: binding the platform's three-dimensional position coordinates, platform velocity vector, platform attitude angle, received signal strength, time difference of arrival, Doppler frequency shift, RF fingerprint features, RF receiver sampling timestamp, and positioning module timing timestamp at the same measurement moment. This binding is performed according to the nearest neighbor time registration between the RF receiver sampling timestamp and the positioning module timing timestamp, ensuring that the RF observation data and platform position data fall within the same measurement moment record. Data records with missing timestamps, missing platform three-dimensional position coordinates, and missing RF measurement fields are removed. Missing RF measurement fields refer to the absence of any one of the following: received signal strength, time difference of arrival, Doppler frequency shift, or RF fingerprint features. Furthermore, duplicate records and records with reversed time sequences between adjacent measurement moments are cleaned up. To avoid repeatedly writing the same spatial location into the subsequent virtual multi-base station candidate observation set, the platform's three-dimensional position coordinates output by the RTK-GNSS positioning module are transformed to a unified local coordinate system, while retaining the coordinate transformation relationship with the WGS84 geodetic coordinate system. The unified local coordinate system is determined by the relative positional relationship between the platform's three-dimensional position coordinates and the coordinate transformation relationship with the WGS84 geodetic coordinate system. This system is used to ensure that each candidate virtual base station participates in joint positioning calculation under the same coordinate reference. Furthermore, Z-score standardization is performed on the received signal strength, time difference of arrival, Doppler frequency shift, platform velocity vector magnitude, and platform attitude angle, so that data with different dimensions are on a unified numerical scale in the subsequent calculation of RF fingerprint matching results, Doppler matching results, track matching results, and observation quality scores.

[0037] In this embodiment, Z-score normalization is also used for numerical anomaly identification. When the normalization results of received signal strength, time difference of arrival, Doppler frequency shift, platform velocity vector magnitude, and platform attitude angle exceed 3 to 4 times the standard deviation, the corresponding data record is removed as a numerical anomaly record. The standard deviation range is determined by the static hovering sampling segment before the start of the mission or by historical data of similar missions.

[0038] This implementation plan establishes a standardized observation data foundation with consistent timing, unified coordinates, valid fields, and comparable numerical scales. This ensures that radio frequency observation data and platform location data can accurately correspond at the same measurement time, reducing interference from missing records, duplicate records, and time-series anomalies in the construction of virtual multi-base station candidate observation sets. It also guarantees that the platform's three-dimensional location coordinates are suitable for local joint positioning calculations and can be stably converted to the WGS84 geodetic coordinate system in the subsequent geographic coordinate mapping output stage. This improves the consistency of calculations for radio frequency fingerprint matching results, Doppler matching results, track matching results, and observation quality scores.

[0039] Specifically, the process of constructing a virtual multi-base station candidate observation set based on UAV maneuvering observation data is as follows: The platform's three-dimensional position coordinates are paired with received signal strength, time difference of arrival, Doppler frequency shift, and radio frequency fingerprint features according to the measurement time. This allows a single UAV to form a single maneuvering observation record defined by both spatial position and radio frequency observation at each measurement time. The platform's three-dimensional position coordinates at each measurement time are designated as a candidate virtual base station. These candidate virtual base stations are used to equivalently represent the spatial receiving positions of the broadband radio frequency receiver at different measurement times. The candidate virtual base stations, along with the received signal strength, time difference of arrival, Doppler frequency shift, and radio frequency fingerprint features at the same measurement time, are written into the virtual multi-base station candidate observation set. The platform's three-dimensional position coordinates at the k-th measurement time are designated as follows: The radio frequency observation data at the k-th measurement time includes received signal strength, time difference of arrival, Doppler frequency shift, and radio frequency fingerprint features. The virtual multi-base station candidate observation set is represented as:

[0040] ;

[0041] In the formula, Represents a virtual multi-base station candidate observation set; The three-dimensional position coordinates of the platform at the k-th measurement time are obtained by the RTK-GNSS positioning module and are used to calibrate the spatial position of the k-th candidate virtual base station. The radio frequency observation data at the k-th measurement time consists of the received signal strength, time difference of arrival, Doppler frequency shift, and radio frequency fingerprint features collected at the same measurement time. This represents the cumulative number of measurements. Each item in the virtual multi-base station candidate observation set corresponds to a candidate observation record at the same measurement time. and Data is associated one by one according to the measurement time, and data records that have not passed the preprocessing are not written into the virtual multi-base station candidate observation set.

[0042] In this implementation plan, a data organization foundation for a virtual multi-base station candidate observation set is established. The spatial receiving position formed by a single UAV at multiple measurement times is uniformly expressed as a candidate virtual base station. Each candidate virtual base station has a clear platform three-dimensional position coordinate and corresponding radio frequency observation data. This provides a calculable object for distinguishing the observation quality and geometric contribution of multi-time slot observation points, and avoids the problem of repeated observations on the same side continuously participating in joint positioning caused by simply accumulating observation data according to the number of measurements.

[0043] Specifically, the process of obtaining a set of candidate observations from the same source, based on RF fingerprint, Doppler, and track matching results, and using observation quality scores generated by time synchronization deviation values ​​to screen candidate virtual base stations, is as follows: RF fingerprint features from adjacent measurement times are matched using spectral peak positions, bandwidth profiles, and energy distribution similarity to obtain RF fingerprint matching results. Specifically, spectral peak positions are matched using peak index alignment, bandwidth profiles using dynamic time warping, and energy distributions using cosine similarity. These three matching results are then normalized and fused into an RF fingerprint matching result. The spectral peak position corresponds to the peak index deviation, the bandwidth profile to the dynamic time warping distance, and the energy distribution to the cosine distance. After normalization, the RF fingerprint matching result is generated according to the rule that the smaller the deviation, the higher the matching result. The Doppler frequency shift, platform velocity vector, and platform three-dimensional position coordinates from adjacent measurement times are subjected to residual verification to obtain Doppler matching results. Specifically, the Doppler frequency shift... The consistency trend of the platform velocity vector projection along the line connecting adjacent candidate virtual base stations is verified, and the verification residual is used as the basis for generating the Doppler matching result. The verification residual is obtained by the absolute value of the difference between the adjacent changes in Doppler frequency shift and the changes in the projection of the platform velocity vector. The smaller the verification residual, the higher the Doppler matching result. The platform's three-dimensional position coordinates, platform velocity vector, and platform attitude angles at the previous measurement time are extrapolated to the current measurement time, and the position deviation is verified with the platform's three-dimensional position coordinates at the current measurement time to obtain the track matching result. The extrapolation process uses the platform velocity vector for position recursion and uses the platform attitude angle to constrain the heading change. The position deviation verification is used to identify candidate virtual base stations caused by track jumps, attitude changes, or platform position anomalies. The position deviation verification is determined by the Euclidean distance between the extrapolated platform three-dimensional position coordinates and the platform's three-dimensional position coordinates at the current measurement time. The smaller the Euclidean distance, the higher the track matching result.

[0044] Whether a candidate observation is retained is determined jointly by radio frequency fingerprint consistency, Doppler consistency, and trajectory prediction consistency, using the following formula:

[0045] ;

[0046] In the formula, The set of joint consensus candidate observations at time k is obtained by the intersection of radio frequency fingerprint consistency, Doppler consistency and track prediction consistency, and is used to determine whether candidate virtual base stations at time k are included in the subsequent observation quality scoring. The set of candidate observations that meet the RF fingerprint consistency requirements at the k-th measurement time is obtained by matching the spectral peak position, bandwidth profile and energy distribution similarity of the RF fingerprint features, and is used to determine whether the RF signals at adjacent measurement times have the same source; The set of candidate observations that meet the Doppler consistency requirement at the k-th measurement time is obtained by checking the residuals between the Doppler frequency shift, the platform velocity vector, and the platform's three-dimensional position coordinates at adjacent measurement times. It is used to determine whether the Doppler frequency shift change conforms to the UAV's motion trend. This represents the candidate observation set that meets the trajectory prediction consistency requirement at the k-th measurement time. It is obtained by verifying the position deviation after extrapolation of the platform's three-dimensional position coordinates, platform velocity vector, and platform attitude angles at the previous measurement time. It is used to exclude candidate virtual base stations that exhibit trajectory jumps or platform positioning anomalies. Only when a candidate virtual base station at the k-th measurement time simultaneously belongs to... , and At that time, the candidate virtual base stations at the k-th measurement time enter the subsequent observation quality scoring determination.

[0047] The absolute value of the difference between the sampling timestamp of the RF receiver and the timing timestamp of the positioning module is used as the time synchronization deviation value. This time synchronization deviation value characterizes the synchronization error between the RF observation data and the platform position data when time-binding. Normalized deduction fusion is performed on the inter-frame jump values ​​of the received signal strength, the residual dispersion of the time difference of arrival, the residual dispersion of the Doppler frequency shift, and the time synchronization deviation value to obtain an observation quality score. This score characterizes the signal stability and time-binding reliability of a single measurement record. Specifically, the normalized deduction fusion incorporates the inter-frame jump values ​​of the received signal strength, the dispersion of the time difference of arrival, and the dispersion of the Doppler frequency shift. Time synchronization deviations are converted into quality deduction items, and an observation quality score is obtained through weighted deduction. Each quality deduction item is normalized to the interval between 0 and 1. The observation quality score is obtained by subtracting the weighted result of each quality deduction item from 1. The observation quality score ranges from 0 to 1; the larger the quality deduction item, the lower the observation quality score. The inter-frame jump value of the received signal strength is obtained from the absolute value of the difference in received signal strength between adjacent measurement times. Outlier determination based on interquartile range is performed on the RF fingerprint matching results, Doppler matching results, track matching results, and observation quality scores of candidate virtual base stations. Outlier determination for quartiles first involves converting RF fingerprint matching results, Doppler matching results, track matching results, and observation quality scores into outlier deviations. Then, the first quartile, third quartile, and interquartile range (IQR) of the four types of outliers are calculated. The IQR is the difference between the third quartile and the first quartile. The main distribution interval is determined by the upper and lower boundaries formed by the first quartile, third quartile, and IQR. The lower boundary of the main distribution interval is obtained by subtracting 1.5 to 3 times the IQR from the first quartile, and the upper boundary is determined by the third quartile. The result is obtained by adding 1.5 to 3 times the interquartile range. In ordinary field inspection scenarios, 1.5 times the interquartile range is used, while in mountainous and forested areas with strong multipath obstruction or radio interference investigation scenarios, 2 to 3 times the interquartile range is used. Candidate virtual base stations that deviate from the main distribution range are judged as outliers. Candidate virtual base stations that do not constitute outliers are retained in the same source candidate observation set, while candidate virtual base stations that constitute outliers are excluded from the data range of the current joint positioning solution, so that the candidate virtual base stations in the same source candidate observation set are consistent in terms of radio frequency characteristics, motion trends, track continuity and observation quality.

[0048] In this implementation plan, a dual screening result based on the homogeneity and observation quality of candidate virtual base stations is formed. This ensures that candidate virtual base stations entering the homogeneous candidate observation set are consistent in terms of radio frequency source, UAV motion trend, platform track continuity, and measurement reliability. Outliers with abnormal radio frequency characteristics, abnormal motion relationships, abnormal platform positions, and poor time-binding quality are excluded. This avoids low-quality multi-time-slot observation points from directly participating in the current joint positioning solution, and provides a stable data foundation for subsequently distinguishing the geometric contribution of candidate virtual base stations, suppressing duplicate observations on the same side, and improving the convergence of target position uncertainty.

[0049] Specifically, the process of constructing a contribution evaluation dataset by combining the candidate observation set from the same source is as follows: The first q candidate virtual base stations are selected from the candidate observation set according to the measurement time. The distance between the platform's three-dimensional position coordinates of any two selected candidate virtual base stations exceeds the larger of the platform positioning accuracy values ​​at the two measurement times. Here, q represents the number of candidate virtual base stations initially written into the selected virtual base station set. q is determined by the number of candidate virtual base stations in the candidate observation set that meet the platform's three-dimensional position coordinate distance requirement. q is a positive integer not less than 2 and not greater than the total number of candidate virtual base stations in the candidate observation set. When the number of selected virtual base stations is less than two, the calculation of geometric angle change is temporarily not performed, so that the selected virtual base station set has an initial observation benchmark for subsequent geometric contribution comparison. Before subsequent candidate virtual base stations enter the contribution judgment, the candidate virtual base stations are merged with the selected virtual base station set to form an observation dataset to be updated. The observation dataset to be updated is used to simulate the impact of writing candidate virtual base stations on the observation geometry and positioning uncertainty, without directly changing the current state of the selected virtual base station set.

[0050] Based on the baseline direction between the candidate virtual base stations and the selected virtual base station set, the change in the minimum included angle of the spatial distribution before and after the addition of the candidate virtual base station is calculated, yielding the geometric angle change. The geometric angle change characterizes the degree to which the addition of the candidate virtual base station alters the spatial angle of the selected virtual base station set. The minimum included angle of the spatial distribution is the smallest angle among the baseline directions formed between the candidate virtual base station and each selected virtual base station. The geometric angle change characterizes whether the candidate virtual base station can break the same-side clustering or near-collinear distribution. Based on the relationship between the candidate virtual base station and the selected virtual base station... The platform's three-dimensional position coordinates, baseline direction angles, and RF fingerprint feature similarity are used to determine distance proximity, directional repetition, and signal feature repetition relationships. These are then normalized and fused according to the degree of in-direction clustering to generate a redundancy level. This redundancy level characterizes the spatial and signal feature repetition between candidate virtual base stations and the selected virtual base station set. Distance proximity is obtained by sorting the Euclidean distance between the platform's three-dimensional position coordinates, directional repetition is obtained by sorting the baseline direction angles, and signal feature repetition is obtained by sorting the RF fingerprint feature similarity. The redundancy level is determined by the distance proximity, directional repetition, and signal feature repetition. The complex relation is obtained by normalization fusion, with values ​​ranging from 0 to 1. The redundancy increases with the proximity of candidate virtual base stations, repetition of directions, and enhancement of signal feature repetition. The observation dataset to be updated is used as the contribution evaluation dataset. The Fisher information matrix is ​​updated based on the observation equation formed by the received signal strength, time difference of arrival, and Doppler frequency shift. The Fisher information matrix is ​​determined by the partial derivative of the observation equation formed by the received signal strength, time difference of arrival, and Doppler frequency shift with respect to the target position x and the corresponding measurement noise, with the target position x as the parameter vector. The posterior covariance before and after combination is calculated. The decrease in the trace of the difference matrix is ​​used to obtain the decrease in posterior covariance. The decrease in posterior covariance is the percentage decrease in the trace of the posterior covariance matrix before and after the candidate virtual base station is added to the observation dataset to be updated. It is obtained by subtracting the trace of the posterior covariance matrix after the addition from the trace of the posterior covariance matrix before the addition, and then dividing by the trace of the posterior covariance matrix before the addition. The decrease in posterior covariance is used to characterize the actual contribution of the candidate virtual base station to reducing the uncertainty of the target position. During the first execution, the coarse positioning position formed by the same source candidate observation set is used as the linearized position for contribution evaluation; during subsequent executions, the target position estimate from the previous round is used as the linearized position for contribution evaluation.

[0051] The cumulative Fisher information matrix corresponding to the contribution evaluation dataset is updated using the cumulative Fisher information matrix format:

[0052] ;

[0053] In the formula, The cumulative Fisher information matrix up to the k-th measurement time is formed by sequentially summing the data from each virtual base station in the contribution assessment dataset. It is used to characterize the overall information content of the current observation data for the target location estimation. The initial information matrix is ​​obtained by setting the uncertainty of the initial target position estimate or inverting the initial covariance matrix, and is used to provide the initial information benchmark for joint positioning solutions. This represents the three-dimensional position coordinates of the i-th selected virtual base station on the platform. At the target position The information matrix increment provided is obtained by differentiating the observation equation formed by the received signal strength, time difference of arrival, and Doppler frequency shift of the i-th selected virtual base station, and is used to characterize the contribution of a single selected virtual base station to reducing the uncertainty of the target location; The three-dimensional position coordinates of the platform at the time of the i-th measurement are obtained by the RTK-GNSS positioning module and are used to determine the spatial location of the i-th selected virtual base station. The target location is represented by the iterative estimation obtained from the joint positioning solution; the posterior covariance matrix is ​​obtained by inverting the cumulative Fisher information matrix. The trace of the posterior covariance matrix is ​​used to characterize the overall uncertainty of the target location estimation. The decrease ratio of the trace of the posterior covariance matrix before and after combination is used as the posterior covariance reduction amount, which is used to determine whether the candidate virtual base station has an actual effect on reducing the positioning uncertainty.

[0054] In this implementation plan, a geometric contribution pre-evaluation mechanism is established before candidate virtual base stations are included in the set of selected virtual base stations. This mechanism ensures that candidate virtual base stations are not directly included in the joint positioning solution simply because they share the same observation source or their measurement quality is up to standard. Instead, their actual effects on the spatial distribution of observations, suppression of repeated observations, and reduction of target location uncertainty are verified in the observation dataset to be updated. Candidate virtual base stations that are clustered on the same side, have a near-collinear distribution, or lack sufficient information increments are screened out. Observation points that can effectively improve the convergence state of the posterior covariance matrix are retained. This provides a contribution screening basis for suppressing the interference of repeated observations in multiple time slots on the joint positioning solution and improving the stability of geographic coordinate output.

[0055] Specifically, the process of selecting virtual base stations by contribution coefficient and identifying the observed geometric degradation state and degradation reinforcement direction is as follows: The geometric angle change, redundancy level, and posterior covariance decrease are input into a contribution discriminant function to generate contribution coefficients. This function determines the contribution coefficients of candidate virtual base stations based on the geometric angle change, redundancy level, and posterior covariance decrease. The normalized geometric angle change, normalized posterior covariance decrease, and redundancy level all range from 0 to 1, while the contribution coefficient ranges from 0 to 1. The contribution coefficient increases with increasing geometric angle change and posterior covariance decrease, and decreases with increasing redundancy level. The contribution threshold is obtained by outlier determination based on interquartile range (ICM) of the contribution coefficients of each candidate virtual base station in the current round of candidate observations. The contribution threshold is taken as the lower boundary of the main distribution interval of the contribution coefficients in the current round of candidate observations, and its range is 0.35 to... The contribution coefficient is 0.65. When the number of selected virtual base stations is small at the beginning of the mission, a value close to 0.35 is used. When the selected virtual base station set has covered multiple spatial orientations, a value close to 0.65 is used. The contribution coefficient is used to simultaneously measure the spatial reinforcement effect, repeated observation suppression effect, and uncertainty reduction effect of the candidate virtual base stations. The contribution discrimination function uses the product term of the geometric angle change and the posterior covariance decrease to characterize the gain contribution. The exponential attenuation term of redundancy is used to suppress the repeated observation contribution. This ensures that even if the radio frequency observation quality is stable, the contribution coefficient of the candidate virtual base stations obtained by repeated measurements on the same side will decrease due to the increased redundancy. When the contribution coefficient reaches the contribution threshold, the candidate virtual base station is added to the selected virtual base station set. When the contribution coefficient does not reach the contribution threshold, the candidate virtual base station is added to the redundant candidate observation record. The redundant candidate observation record retains the measurement time and the three-dimensional position coordinates of the platform for subsequent field playback analysis but does not participate in the current joint positioning solution.

[0056] like Figure 2The diagram shows the flowchart for observational geometric degradation identification and degradation reinforcement direction generation. A spatial distribution matrix is ​​constructed based on the 3D position coordinates of the platform in the selected virtual base station set. Principal component analysis is used to obtain the first direction eigenvalue, the second direction eigenvalue, and the main distribution direction of the base stations. The first direction eigenvalue represents the dispersion of the selected virtual base station set along the main distribution direction, and the second direction eigenvalue represents the dispersion of the selected virtual base station set along the vertical direction. The posterior covariance matrix obtained by inverting the Fisher information matrix during the first observational geometric degradation identification, or the previous round posterior covariance matrix read during subsequent observational geometric degradation identification, is subjected to eigenvalue decomposition. The direction of the largest eigenvalue is taken as the major axis direction of the posterior covariance ellipse, which represents the direction with the greatest uncertainty in the current target position estimation. When the ratio of the first direction eigenvalue to the second direction eigenvalue reaches the geometric degradation judgment threshold, and the base station... When the angle between the main distribution direction and the major axis of the posterior covariance ellipse is not greater than the degradation angle threshold, the current observation geometry is determined to be in a degraded state. Among the candidate normal directions on both sides of the major axis of the posterior covariance ellipse, the direction that maximizes the decrease in the predicted posterior covariance matrix trace is selected as the degradation reinforcement direction, constraining the next optimal observation waypoint to move towards the spatial orientation that reduces the uncertainty of the major axis of the posterior covariance ellipse. Otherwise, the current observation geometry is determined to be in a non-degraded state, and the degradation reinforcement direction is set to empty. The value range of the geometry degradation judgment threshold is 3 to 5, and the value range of the degradation angle threshold is 15° to 30°. The geometry degradation judgment threshold and the degradation angle threshold are determined by the eigenvalue ratio and direction angle statistics of the non-degraded observation geometry completed in the task area. When the selected virtual base station set is concentrated along a single-sided track, the geometry degradation judgment threshold is taken as close to 3, and the degradation angle threshold is taken as close to 30°.

[0057] This implementation plan establishes a contribution admission and geometric degradation diagnosis mechanism for the selected virtual base station set. This mechanism determines whether a candidate virtual base station participates in the current joint positioning solution based on its contribution coefficient, mitigating the impact of identical-side repeated measurements, nearest-neighbor repeated measurements, and insufficient information increments on the positioning results. It retains candidate virtual base stations that provide spatial reinforcement and uncertainty reduction. Simultaneously, it identifies whether the selected virtual base station set exhibits geometric degradation states such as identical-side clustering, near-collinear distribution, or unconstrained major axis directions of the posterior covariance ellipse. This generates degradation reinforcement directions, providing a clear geometric reinforcement basis for the next optimal waypoint planning, and improving the problems of insufficient convergence of target position uncertainty and poor stability of geographic coordinate output.

[0058] Specifically, the process of constructing three types of observation constraints based on the selected virtual base station set is as follows: First, the received signal strength and platform 3D position coordinates are converted into a received signal strength ranging constraint according to the free space path loss relationship. The free space path loss relationship uses the attenuation relationship of the received signal strength as the target distance changes to form a distance observation equation. Second, the arrival time difference and platform 3D position coordinates are converted into a hyperbolic time difference constraint according to the distance difference relationship. The distance difference relationship uses the first selected measurement time as a time reference to form a hyperbolic observation equation. The hyperbolic time difference constraint is constructed from the arrival time difference between the corresponding measurement time of each selected virtual base station and the first selected measurement time. Third, the Doppler frequency shift, platform velocity vector, and platform 3D position coordinates are converted into a Doppler frequency shift azimuth constraint according to the velocity projection relationship. The velocity projection relationship uses the projection change of the platform velocity vector in the target azimuth direction to determine the azimuth constraint. The target azimuth direction is determined by the direction vector pointing from the platform 3D position coordinates to the target position. The Doppler frequency shift azimuth constraint is formed by the relationship between the projection of the platform velocity vector on this direction vector and the carrier frequency.

[0059] The received signal strength ranging constraint adopts a free space path loss model:

[0060] ;

[0061] ;

[0062] In the formula, The received signal strength at the k-th measurement time is directly acquired by the wideband radio frequency receiver at the k-th measurement time and is used as the observation input for the received signal strength ranging constraint. The transmit power is indicated by the known transmit parameters obtained from the radio management database, the technical specifications of the transmitting equipment model, the field calibration parameters, or the offline fingerprint database, and is used to determine the transmit power reference of the radio frequency signal source. This indicates the transmit antenna gain, which is obtained from the technical specifications of the transmitting equipment or field calibration parameters, and is used to correct the influence of the transmit antenna directivity on the received signal strength. This indicates the receiving antenna gain, which is obtained from the factory calibration curve of the wideband RF receiver or the calibration results of the standard signal source, and is used to correct for differences in antenna gain on the receiving side. This represents the wavelength corresponding to the carrier frequency, which is calculated from the monitoring center frequency and is used to characterize the wavelength parameter of radio frequency signals propagating in free space. This represents the distance between the target position and the platform's three-dimensional position coordinates at the k-th measurement time, determined by the target position. platform three-dimensional position coordinates The Euclidean distance between them is obtained and used to convert the received signal strength into a distance constraint; This represents additional propagation loss, estimated by empirical models of terrain shading and vegetation, or obtained by online correction of the residuals of received signal strength at multiple times, used to compensate for environmental attenuation beyond free-space propagation; The measurement noise is determined by the residual of the received signal strength and the noise level of the wideband RF receiver, and is used to characterize the power measurement disturbance at the k-th measurement time. The target location is represented by iterative estimation obtained from the joint positioning solution. The three-dimensional position coordinates of the platform at the k-th measurement time are obtained by the RTK-GNSS positioning module and are used to calibrate the spatial position of the k-th selected virtual base station. When the transmit power, transmit antenna gain, or additional propagation loss cannot be directly obtained, the corresponding parameters are used as parameters of the model to be calibrated to participate in the joint positioning solution. The received signal strength ranging constraint is directly established when the transmit parameters, receive antenna gain, and corresponding wavelength of the carrier frequency are known. When the transmit power, transmit antenna gain, or additional propagation loss is missing, the parameters of the model to be calibrated participate in the joint positioning solution.

[0063] In this embodiment, additional propagation loss The loss is obtained by adding vegetation crossing loss, terrain shading loss, and online correction loss of received signal strength residuals. Vegetation crossing loss is obtained by multiplying the length of the vegetation area through which the line connecting the candidate virtual base station to the estimated target location passes by the unit vegetation loss coefficient, with the engineering value of the unit vegetation loss coefficient ranging from 3dB to 8dB per 100 meters. Terrain shading loss is determined by the difference in shading height between the line connecting the candidate virtual base station to the estimated target location and the terrain profile, with the engineering value of the terrain shading loss ranging from 5dB to 20dB. The online correction loss of received signal strength residuals is updated by updating the average value of received signal strength residuals at multiple consecutive measurement times.

[0064] Based on the time synchronization deviation, the change in platform velocity vector direction, and the outlier results of the observation residuals of the three types of observation constraints obtained from the previous round of joint positioning solution or the current round of pre-solution, the three types of observation constraints are set to an enabled state, a reduced-weight state, or a suspended state, respectively, as their participation states. Specifically, the time synchronization deviation is used to adjust the participation state of the hyperbolic time difference constraint, the change in platform velocity vector direction is used to adjust the participation state of the Doppler frequency shift azimuth constraint, and the outlier results of the observation residuals are used to adjust the participation state of the three types of observation constraints in the current round of joint positioning solution. When the corresponding adjustment basis does not constitute an outlier observation, it is set to an enabled state; when the corresponding adjustment basis constitutes an outlier observation but still retains the same source candidate observation relationship, it is set to a reduced-weight state; when the corresponding adjustment basis constitutes an outlier observation and the outlier results of the observation residuals of the corresponding observation constraints are also abnormal, it is set to a suspended state. The fusion weight in the reduced-weight state is lower than the fusion weight in the enabled state, and the corresponding observation constraint in the suspended state does not enter the current round of joint positioning solution.

[0065] In this implementation plan, a multi-constraint joint positioning foundation is formed corresponding to the selected virtual base station set. By participating in state control, various observation constraints are prevented from being mechanically incorporated into the current joint positioning solution under unsuitable conditions. This allows the received signal strength ranging constraint, time difference of arrival hyperbolic constraint, and Doppler frequency shift azimuth constraint to be enabled, downweighted, or suspended according to the actual observation reliability. This reduces the interference of time synchronization deviation, platform motion direction changes, and observation residual outliers on the target position estimate, improves the ability of the posterior covariance matrix to represent the true positioning uncertainty, and provides an adaptive processing basis at the constraint level for suppressing the continuous impact of low-quality observations and insufficiently contributing observations on the joint positioning solution and improving the stability of geographic coordinate output.

[0066] Specifically, the process of generating target position estimates, posterior covariance matrices, non-line-of-sight high-risk sectors, and the next optimal waypoint through joint positioning is as follows: Observation constraints entering the joint positioning solution are selected according to their participation status among the three types of observation constraints. Fusion weights are determined based on the traces of the posterior covariance matrices of the three types of observation constraints, resulting in the target position estimate and posterior covariance matrix. The traces of the posterior covariance matrices of the three types of observation constraints are obtained by each corresponding observation constraint participating individually in the current round of pre-solution. The fusion weights are proportional to the reciprocal of the traces of the posterior covariance matrices and normalized until the sum of all fusion weights is 1. Observation constraints in the active state enter the joint positioning solution; observation constraints in the deweighted state enter the joint positioning solution with fusion weights lower than those in the active state; and observation constraints in the suspended state do not enter the current round of joint positioning solution. Non-line-of-sight observations are identified based on the observation residuals obtained from the joint positioning solution, and n consecutive measurements are... During the measurement process, the observation residuals of the received signal strength ranging constraint and the observation residuals of the arrival time difference hyperbolic constraint both constitute the selected virtual base stations for positive outlier observations. These are connected to the target position estimate to calculate the azimuth angle of the outlier observations. Based on the coverage of the azimuth angle of the outlier observations, non-line-of-sight high-risk sectors are generated to record the spatial azimuth range of continuous positive outlier observations and serve as avoidance constraints when generating the next optimal waypoint. Positive outlier observations refer to observations with positive observation residuals that deviate from the upper boundary of the main distribution interval obtained by outlier determination based on interquartile range. Non-line-of-sight high-risk sectors are used to record the spatial azimuth range of continuous positive outlier observations, transforming outlier observations from single residual processing into constraints for the next waypoint planning. Here, n is the continuous observation window length, which is obtained by rounding up the ratio of the minimum duration of continuous anomalies to the adjacent time interval of the RF receiver sampling timestamp, and n is a positive integer not less than 2.

[0067] The positive distance offset relationship for non-line-of-sight observations adopts a non-line-of-sight measurement model:

[0068] ;

[0069] In the formula, The distance observation value at the k-th measurement time is calculated by the received signal strength ranging constraint based on the received signal strength and the three-dimensional position coordinates of the platform, and is used to characterize the apparent target distance at the current measurement time. The distance value at the k-th measurement time under line-of-sight propagation conditions is obtained from the Euclidean distance between the target position estimate and the platform's three-dimensional position coordinates at the k-th measurement time, and is used as a reference distance under conditions without diffraction obstruction. This represents the positive distance offset introduced at the k-th measurement time under non-line-of-sight propagation conditions, by... and The difference is determined to characterize the extent to which occlusion or multipath propagation causes the distance observation to be overestimated, and serves as the basis for the observation residuals constrained by the received signal strength ranging to constitute positive outlier observations.

[0070] Candidate waypoints are generated based on the major axis direction of the posterior covariance ellipse, the degradation state of the observation geometry, the degradation reinforcement direction, and non-line-of-sight (NLS) high-risk sectors. The candidate waypoint set consists of discrete waypoints within flyable areas that satisfy maximum flight speed, minimum safe flight altitude, and airspace control constraints. These waypoints are generated discretely within the single-segment reachability range of the current waypoint, according to their bearing and distance. Candidate waypoints located within NLS high-risk sectors are excluded. When the current observation geometry is in a degradation state, waypoints are sorted in ascending order of their angle with the degradation reinforcement direction. This sorting prioritizes the next optimal waypoint located where reinforcement is possible. The spatial orientation of the major axis of the posterior covariance ellipse is determined; for the ranked candidate observation waypoints, the decrease in the trace of the posterior covariance matrix caused by the formation of new candidate virtual base stations and their addition to the set of selected virtual base stations is calculated, as well as the turning angle and range increment generated when flying from the current waypoint to the candidate observation waypoint. The next optimal observation waypoint and the flight path of the next segment are then determined. The next optimal observation waypoint is determined according to the comprehensive ranking results of the decrease in the trace of the posterior covariance matrix from large to small and the turning angle and range increment from small to large. The flight path of the next segment avoids high-risk sectors outside line of sight and takes into account observation geometric reinforcement and flight cost constraints.

[0071] In this embodiment, as Figure 3The diagram illustrates the identification of observation geometry degradation and the reinforcement of reverse waypoints. In the diagram, black "×" represents the estimated target position; the red dashed ellipse represents the posterior covariance ellipse, and the red arrow indicates the direction of the major axis of the posterior covariance ellipse, used to indicate the direction of the greatest current positioning uncertainty; gray hollow circles represent redundant candidate observation records; "Same-side clustering, insufficient angular extension" indicates that candidate virtual base stations are concentrated on the same side, making it difficult to effectively improve the observation geometry; blue square dots represent the selected set of virtual base stations; light red shaded sectors represent high-risk non-line-of-sight sectors, and "× avoid" indicates that subsequent candidate waypoints must avoid this sector; green dotted arrows indicate the direction of degradation reinforcement, and green pentagrams indicate the next optimal observation waypoint; orange triangles represent the current waypoint, and purple dashed lines represent the flight path for the next segment. As can be seen from the diagram, this scheme can guide the UAV to avoid high-risk non-line-of-sight sectors and plan the next optimal observation waypoint in the direction of degradation reinforcement after identifying same-side redundant observations and observation geometry degradation.

[0072] In this embodiment, as Figure 4 The diagram shows the comparison results of the engineering verification. A ground radio frequency signal source was set up in the 2000m × 2000m mountainous forest area mission area to control a six-rotor industrial UAV to complete the initial maneuver observation along the same side track. Two processing methods were compared: the first processing method did not perform contribution coefficient screening, did not generate degradation reinforcement direction, and did not exclude non-line-of-sight high-risk sectors. The second processing method performed the same source candidate observation set screening, contribution coefficient screening, observation geometric degradation state identification, non-line-of-sight high-risk sector generation, and next optimal observation waypoint planning according to this embodiment. Figure 4 The gray rectangular area represents the range of values ​​for the circular probability error (CEP) and root mean square (RMS) localization error corresponding to the first processing method, while the blue rectangular area represents the range of values ​​for the circular probability error (CEP) and RMS localization error corresponding to the second processing method. The dots within the rectangular areas represent the center positions of the intervals for the circular probability error (CEP) and RMS localization error under the corresponding processing method. The elliptical lines represent the elliptical parameters of the target position covariance under the corresponding processing method, and the arrows indicate the direction of error reduction from the first processing method to the second processing method. Through multiple sets of initial trajectory simulation comparisons, the circular probability error (CEP), root mean square (RMS) positioning error, and the length of the major axis of the target position covariance ellipse were all lower in the second processing method than in the first processing method. In an engineering verification result, the RMS positioning error decreased from 38-55 meters to 18-30 meters, the CEP decreased from 28-42 meters to 14-25 meters, and the length of the major axis of the target position covariance ellipse decreased by 35% to 55%. This indicates that the contribution coefficient screening and degradation reinforcement direction can reduce the impact of repeated observations on the same side on the joint positioning solution.

[0073] In this embodiment, the method is applied to scenarios such as emergency search and rescue in mountainous areas, radio interference investigation, and field IoT device inspection. In the emergency search and rescue scenario, the ground radio frequency signal source is a low-power emergency beacon, and a six-rotor industrial UAV performs mobile radio frequency observations in areas obscured by tree canopies and undulating terrain. In the radio interference investigation scenario, the ground radio frequency signal source is an abnormal transmitting device, and non-line-of-sight high-risk sectors are used to constrain the flight path of the next flight segment to avoid continuous positive outlier observation directions. In the field IoT device inspection scenario, the ground radio frequency signal source is the IoT terminal to be inspected, and the next optimal observation waypoint is used to reinforce the uncertain region corresponding to the major axis direction of the posterior covariance ellipse. In the above scenarios, the airborne edge computing unit completes the step processing locally on the UAV and outputs WGS84 geographic coordinates, altitude, and positioning accuracy information, enabling the method to meet the engineering application requirements of rapid search, stable positioning, and geographic coordinate mapping of radio frequency signal sources under complex field conditions.

[0074] In this implementation plan, a feedback loop is formed between the joint positioning solution and the next waypoint planning. This ensures that the target position estimate and the posterior covariance matrix are no longer just used as the current round of positioning output, but are further transformed into the planning basis for non-line-of-sight high-risk sectors, candidate observation waypoint sets, the next optimal observation waypoint, and the flight path of the next flight segment. This avoids repeated observations from continuous abnormal bearings and bearings with low effective contribution, guides newly added candidate virtual base stations to prioritize the reinforcement of uncertain regions corresponding to the major axis of the posterior covariance ellipse, improves the geometric distribution effectiveness of the selected virtual base station set, and addresses the problems of insufficient convergence of target position uncertainty and poor stability of geographic coordinate output.

[0075] Specifically, the process of determining whether to terminate maneuvering radio frequency observation by judging the decreasing relationship of positioning posterior variance, geometric accuracy factor, normalized residual mean square, and new observation sampling gain, and outputting WGS84 geographic coordinates, altitude, and positioning accuracy information after the acquisition is terminated, is as follows: Based on the posterior covariance matrix, the platform's three-dimensional position coordinates in the selected virtual base station set, the observation residual, and the change in the posterior covariance matrix trace before and after adding the next optimal observation waypoint to the selected virtual base station set, the positioning posterior variance, geometric accuracy factor, normalized residual mean square, and new observation sampling gain are obtained respectively. The posterior variance is obtained from the trace of the posterior covariance matrix. The observation geometry matrix is ​​composed of the directional relationship between the platform's three-dimensional position coordinates of each selected virtual base station and the estimated target position. The geometric precision factor is determined by the observation geometry matrix formed by the platform's three-dimensional position coordinates of the selected virtual base station set and the estimated target position, specifically by taking the square root of the trace after inverting the transpose of the observation geometry matrix. A larger geometric precision factor indicates that the spatial distribution of the selected virtual base station set is more likely to cause observation geometry degradation. The normalized residual mean square is obtained by taking the mean square after scaling the observation residuals. The observation residuals originate from the received signal strength entering the joint positioning solution in the current round. The system incorporates ranging constraints, hyperbolic time-of-arrival (TOA) constraints, and Doppler shift azimuth constraints. Scale normalization is performed based on the measurement noise corresponding to the observation constraints. The gain from adding a new observation sampling point is obtained by the change in the posterior covariance matrix trace before and after adding the next optimal waypoint. Specifically, it is calculated by subtracting the posterior covariance matrix trace after adding the waypoint from the trace before adding it, and then dividing by the posterior covariance matrix trace before adding it. If the posterior variance, geometric precision factor, and mean square of normalized residuals do not satisfy a decreasing relationship within m consecutive positioning rounds, or if the gain from adding a new observation sampling point is not higher than the gain from adding a new observation sampling point in the previous round for m consecutive rounds, then the maneuvering radio frequency is terminated. Observe; otherwise, continue collecting data at the next optimal observation waypoint, so that the data collection process is simultaneously constrained by the convergence of positioning uncertainty, geometric convergence of observation, convergence of residual stability, and convergence of new sampling benefits. The positioning solution round refers to a closed-loop solution process that completes a joint positioning solution based on the current set of selected virtual base stations and generates the positioning posterior variance, geometric accuracy factor, normalized residual mean square, and new observation sampling benefits. Pre-solutions that do not generate index sequences are not included in the positioning solution round. m represents the number of continuous convergence comparison rounds, which comes from the number of completed positioning solution rounds. m is a positive integer not less than 2 and m is not greater than the number of currently completed positioning solution rounds.

[0076] After acquisition is terminated, the target position estimate is transformed from the local coordinate system to the WGS84 geodetic coordinate system. The coordinate transformation process retains the transformation relationship between the local coordinate system and the WGS84 geodetic coordinate system established by the RTK-GNSS positioning module. According to this transformation relationship, the target position estimate is reverse-mapped from the unified local coordinate system to the WGS84 geodetic coordinate system. The WGS84 geographic coordinates, altitude, circular probability error (CEP), root mean square error (RMS), and target position covariance elliptic parameters of the ground radio frequency signal source are output. The CEP, RMS, and target position covariance elliptic parameters are used together to characterize the location reliability and directional uncertainty of the geographic coordinate mapping result. The CEP is obtained by converting the plane positioning standard deviation corresponding to the posterior covariance matrix, the RMS is obtained by taking the square root of the square mean of the observation residuals, and the target position covariance elliptic parameters are obtained from the eigenvalues ​​and eigenvectors of the posterior covariance matrix.

[0077] In this implementation plan, a convergence termination and geographic coordinate mapping output mechanism for mobile radio frequency observations is established. This mechanism ensures that the decision to continue data acquisition is no longer solely determined by the number of measurements, but rather by the effectiveness of the current set of virtual base stations, which is jointly reflected by the posterior variance of the location, the geometric accuracy factor, the normalized mean square of the residuals, and the gains from new observation sampling. This avoids the continued introduction of repetitive or low-gain observations on the same side when the uncertainty of the target location is difficult to reduce further, reduces the disturbance of invalid data acquisition to the joint positioning solution results, and outputs WGS84 geographic coordinates with location confidence and directional uncertainty descriptions after the acquisition is terminated, thereby improving the stability and interpretability of the geographic coordinate mapping results.

[0078] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0079] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for RF signal source localization and geographic coordinate mapping based on a single UAV platform, characterized in that, Includes the following steps: S1, during the maneuvering flight of a single UAV, collects and preprocesses the UAV maneuvering observation data; S2. A virtual multi-base station candidate observation set is constructed based on UAV mobile observation data; candidate virtual base stations are screened based on radio frequency fingerprint, Doppler and track matching results, as well as the observation quality score generated by time synchronization deviation value, to obtain a homogeneous candidate observation set; S3, construct a contribution evaluation dataset by combining the candidate observation set from the same source; select the set of virtual base stations by contribution coefficient, and identify the observation geometric degradation state and degradation reinforcement direction; The specific process of selecting the virtual base station set based on the contribution coefficient and identifying the observed geometric degradation state and degradation reinforcement direction is as follows: The geometric angle change, redundancy level, and posterior covariance decrease are input into the contribution discrimination function to generate contribution coefficients. The contribution discrimination function is a function that determines the contribution coefficients of candidate virtual base stations based on the geometric angle change, redundancy level, and posterior covariance decrease. The contribution discrimination function uses the product of the change in geometric angle and the decrease in posterior covariance to characterize the gain contribution, and uses an exponential decay term of redundancy to suppress the contribution of repeated observations. When the contribution coefficient reaches the contribution threshold, the candidate virtual base station is added to the set of selected virtual base stations; When the contribution coefficient does not reach the contribution threshold, the candidate virtual base station is written into the redundant candidate observation record; A spatial distribution matrix is ​​constructed based on the three-dimensional position coordinates of the platform in the selected virtual base station set. Principal component analysis is used to obtain the first directional eigenvalue, the second directional eigenvalue, and the main distribution direction of the base station. The posterior covariance matrix obtained by inverting the Fisher information matrix during the first observation geometry degradation identification, or the previous posterior covariance matrix read during non-first observation geometry degradation identification, is subjected to eigenvalue decomposition. The direction with the largest eigenvalue is taken as the major axis direction of the posterior covariance ellipse. When the ratio of the first directional eigenvalue to the second directional eigenvalue reaches the geometry degradation judgment threshold, and the angle between the main distribution direction of the base station and the major axis direction of the posterior covariance ellipse is not greater than the degradation angle threshold, the current observation geometry is determined to be in a degraded state. Among the candidate normal directions on both sides of the major axis direction of the posterior covariance ellipse, the direction that makes the largest decrease in the predicted posterior covariance matrix trace is selected as the degradation reinforcement direction, which constrains the next optimal observation waypoint to move towards the spatial orientation that reduces the uncertainty of the major axis direction of the posterior covariance ellipse. Otherwise, the current observed geometry is determined to be in a non-degenerate state, and the degenerate reinforcement direction is set to empty; S4, constructing three types of observation constraints based on the selected set of virtual base stations; The target position estimate, posterior covariance matrix, non-line-of-sight high-risk sector and next optimal waypoint are generated by joint positioning solution; S5 determines whether to terminate the mobile radio frequency observation by judging the diminishing relationship between the post-positioning variance, geometric accuracy factor, normalized residual mean square, and new observation sampling gains, and outputs WGS84 geographic coordinates, altitude, and positioning accuracy information after the acquisition is terminated.

2. The method for radio frequency signal source localization and geographic coordinate mapping based on a single UAV platform according to claim 1, characterized in that: The specific process for collecting UAV maneuver observation data during single UAV maneuver flight is as follows: The single UAV platform is equipped with a wideband RF receiver, an RTK-GNSS positioning module, an IMU inertial measurement unit, and an onboard edge computing unit. It controls a single UAV to reach different spatial positions along a planar flight trajectory at multiple measurement moments. During the single UAV's maneuvering flight, it collects UAV maneuvering observation data, which includes: received signal strength, time difference of arrival, Doppler frequency shift, RF fingerprint features, and RF receiver sampling timestamps collected by the wideband RF receiver; platform three-dimensional position coordinates, platform velocity vector, positioning module timing timestamp, and platform positioning accuracy values ​​collected by the RTK-GNSS positioning module; and platform attitude angles collected by the IMU inertial measurement unit.

3. The method for radio frequency signal source localization and geographic coordinate mapping based on a single UAV platform according to claim 1, characterized in that: The specific process of preprocessing is as follows: The platform's three-dimensional position coordinates, platform velocity vector, platform attitude angle, received signal strength, time difference of arrival, Doppler frequency shift, radio frequency fingerprint features, radio frequency receiver sampling timestamp, and positioning module timing timestamp are time-bound at the same measurement time. Data records with missing timestamps, missing platform three-dimensional position coordinates, and missing radio frequency measurement fields are removed. The platform's three-dimensional position coordinates are transformed to a unified local coordinate system, and the received signal strength, time difference of arrival, Doppler frequency shift, platform velocity vector magnitude, and platform attitude angle are Z-score standardized.

4. The method for radio frequency signal source localization and geographic coordinate mapping based on a single UAV platform according to claim 1, characterized in that: The specific process of constructing a virtual multi-base station candidate observation set based on UAV mobile observation data is as follows: According to the measurement time, the platform's three-dimensional position coordinates are paired with the received signal strength, time difference of arrival, Doppler frequency shift, and radio frequency fingerprint features. The platform's three-dimensional position coordinates at each measurement time are designated as a candidate virtual base station. The candidate virtual base stations, along with the received signal strength, time difference of arrival, Doppler frequency shift, and radio frequency fingerprint features at the same measurement time, are written into the virtual multi-base station candidate observation set.

5. The method for radio frequency signal source localization and geographic coordinate mapping based on a single UAV platform according to claim 1, characterized in that: The specific process for obtaining the candidate observation set based on radio frequency fingerprinting, Doppler and track matching results, and observation quality scores generated by time synchronization deviation values ​​to screen candidate virtual base stations is as follows: The RF fingerprint features of adjacent measurement times are matched by spectral peak position, bandwidth profile and energy distribution similarity to obtain RF fingerprint matching results; the Doppler frequency shift, platform velocity vector and platform three-dimensional position coordinates of adjacent measurement times are checked by residual to obtain Doppler matching results; the platform three-dimensional position coordinates, platform velocity vector and platform attitude angle of the previous measurement time are extrapolated to the current measurement time and the position deviation is checked with the platform three-dimensional position coordinates of the current measurement time to obtain track matching results. The absolute value of the difference between the sampling timestamp of the radio frequency receiver and the timing timestamp of the positioning module is used as the time synchronization deviation value. The inter-frame jump value of the received signal strength, the residual dispersion of the time difference of arrival, the residual dispersion of the Doppler frequency shift, and the time synchronization deviation value are normalized and fused to obtain the observation quality score. The observation quality score characterizes the signal stability and time binding reliability of a single measurement record. Outlier determination based on interquartile range is performed on the radio frequency fingerprint matching results, Doppler matching results, track matching results and observation quality scores of candidate virtual base stations. Candidate virtual base stations that do not constitute outlier observations are retained in the same source candidate observation set, while candidate virtual base stations that constitute outlier observations are excluded from the data range of the current joint positioning solution.

6. The method for radio frequency signal source localization and geographic coordinate mapping based on a single UAV platform according to claim 1, characterized in that: The specific process of constructing the contribution evaluation dataset by combining the set of candidate observations from the same source is as follows: Based on the measurement time, the first q candidate virtual base stations are selected from the candidate observation set of the same source. The distance between the platform three-dimensional position coordinates of any two selected candidate virtual base stations exceeds the larger value of the platform positioning accuracy values ​​at the two measurement times. These are then written into the set of selected virtual base stations. Before subsequent candidate virtual base stations enter the contribution judgment, the candidate virtual base stations are merged with the set of selected virtual base stations to form the observation dataset to be updated. Based on the baseline direction between the candidate virtual base station and the platform's three-dimensional position coordinates in the selected virtual base station set, the change in the minimum included angle of the spatial distribution before and after the addition of the candidate virtual base station is calculated to obtain the geometric angle change. The geometric angle change is used to characterize the degree of change in the spatial angle of the selected virtual base station set after the addition of the candidate virtual base station. Based on the platform three-dimensional position coordinate distance, baseline direction angle and radio frequency fingerprint feature similarity between candidate virtual base stations and selected virtual base stations, the distance proximity relationship, direction repetition relationship and signal feature repetition relationship are determined, and normalized fusion is performed according to the degree of same-direction aggregation to generate redundancy degree, which is used to characterize the spatial repetition and signal feature repetition degree between the candidate virtual base station and the set of selected virtual base stations. The observation dataset to be updated is used as the contribution evaluation dataset. The Fisher information matrix is ​​updated based on the observation equation formed by the received signal strength, time difference of arrival, and Doppler frequency shift. The decrease ratio of the trace of the posterior covariance matrix before and after the combination is calculated to obtain the decrease in posterior covariance.

7. The method for radio frequency signal source localization and geographic coordinate mapping based on a single UAV platform according to claim 1, characterized in that: The specific process for constructing the three types of observation constraints based on the selected set of virtual base stations is as follows: Based on the selected set of virtual base stations, three types of observation constraints are constructed: the received signal strength and the platform's three-dimensional position coordinates are converted into a received signal strength ranging constraint according to the free space path loss relationship; the arrival time difference and the platform's three-dimensional position coordinates are converted into an arrival time difference hyperbolic constraint according to the distance difference relationship; and the Doppler frequency shift, the platform velocity vector, and the platform's three-dimensional position coordinates are converted into a Doppler frequency shift azimuth constraint according to the velocity projection relationship. Based on the time synchronization deviation value, the change in the platform velocity vector direction, and the outlier results of the observation residuals of the three types of observation constraints obtained from the previous round of joint positioning calculation or the current round of pre-calculation, the three types of observation constraints are set to the enabled state, the reduced weight state, or the suspended state, respectively, as the participation state of the three types of observation constraints.

8. The method for radio frequency signal source localization and geographic coordinate mapping based on a single UAV platform according to claim 7, characterized in that: The specific process of generating the target position estimate, posterior covariance matrix, non-line-of-sight high-risk sector, and next optimal waypoint through joint positioning calculation is as follows: The observation constraints are selected for joint positioning solution based on their participation status among the three types of observation constraints. The fusion weights are determined based on the traces of the posterior covariance matrices of the three types of observation constraints, resulting in the target position estimate and the posterior covariance matrix. Non-line-of-sight observations are identified based on the observation residuals obtained from the joint positioning solution. The observation residuals of the received signal strength ranging constraint and the observation residuals of the time difference hyperbolic constraint are used to construct virtual base stations for positive outlier observations in n consecutive measurement times. These virtual base stations are connected to the target position estimate to calculate the azimuth angle of the abnormal observations. Based on the coverage of the abnormal observation azimuth angle, a non-line-of-sight high-risk sector is generated to record the spatial azimuth range of continuous positive outlier observations. A set of candidate waypoints is generated based on the major axis direction of the posterior covariance ellipse, the observation geometry degradation state, the degradation reinforcement direction, and the non-line-of-sight high-risk sector. Candidate waypoints located within high-risk sectors outside of line of sight are excluded, and when the current observation geometry is in a degraded state, they are sorted in ascending order of the angle between them and the degradation reinforcement direction; For the sorted candidate waypoints, calculate the decrease in the trace of the posterior covariance matrix caused by the formation of new candidate virtual base stations and their addition to the set of selected virtual base stations, as well as the turning angle and range increment generated when flying from the current waypoint to the candidate waypoint, and determine the next optimal waypoint and the flight path for the next segment.

9. The method for radio frequency signal source localization and geographic coordinate mapping based on a single UAV platform according to claim 8, characterized in that: The specific process of determining whether to terminate mobile radio frequency observation by judging the decreasing relationship between the post-positioning variance, geometric accuracy factor, normalized residual mean square, and new observation sampling gain, and outputting WGS84 geographic coordinates, altitude, and positioning accuracy information after the acquisition is terminated is as follows: Based on the posterior covariance matrix, the platform's 3D position coordinates in the selected virtual base station set, the observation residuals, and the change in the posterior covariance matrix before and after adding a new candidate virtual base station to the selected virtual base station set using the next optimal waypoint, the posterior variance of positioning, geometric precision factor, normalized residual mean square, and new observation sampling gain are obtained respectively. If the posterior variance of positioning, geometric precision factor, and normalized residual mean square do not satisfy the decreasing relationship within m consecutive positioning solution rounds, or if the new observation sampling gain is not higher than the new observation sampling gain in the previous round for m consecutive rounds, then the maneuvering radio frequency observation is terminated. Otherwise, continue data collection at the next optimal waypoint; After the acquisition is terminated, the target position estimate is converted from the local coordinate system to the WGS84 geodetic coordinate system, and the WGS84 geographic coordinates, altitude and positioning accuracy information of the ground radio frequency signal source are output. The positioning accuracy information includes circular probability error (CEP), root mean square positioning error (RMS) and target position covariance ellipse parameters.

Images