Non-cooperative unmanned aerial vehicle pilot identity tracing method based on vehicle space-time correlation

Abstract

The application discloses a non-cooperative unmanned aerial vehicle pilot identity tracing method based on vehicle space-time correlation, comprising the following steps: acquiring feature point data of a target aircraft and forming a point list with space-time attributes; based on a vehicle big data system, screening target vehicle cameras according to the spatial position correlation of each target feature point, and screening license plate number data according to the time correlation to obtain a license plate number array corresponding to each target feature point; processing the license plate number array to obtain a target list sorted by the cross-array appearance frequency of license plate numbers; screening a suspected license plate based on the target list, and performing vehicle and personnel correlation analysis taking the suspected license plate as a verification object to determine a suspected person and complete identity verification. The method can establish a space-time correlation chain between the target aircraft, the ground vehicle and the associated personnel, thereby improving the non-cooperative unmanned aerial vehicle pilot identity tracing capability.

Description

Technical Field

[0001] This application relates to the field of low-altitude safety supervision technology, and in particular to a non-cooperative drone pilot identification method based on vehicle spatiotemporal correlation. Background Technology

[0002] With the rapid development of low-altitude aircraft technologies such as drones and electric vertical takeoff and landing (eVTOL) aircraft, the application of low-altitude airspace in scenarios such as logistics distribution, emergency rescue, security patrols, and urban transportation is becoming increasingly widespread, and the low-altitude economy has become an important development direction. To ensure the safe and orderly conduct of low-altitude flight activities, it is necessary to effectively perceive, monitor, and manage low-altitude flight targets, especially to promptly detect and verify illegal flight behaviors that affect public safety, in order to improve low-altitude safety supervision capabilities.

[0003] Currently, low-altitude airspace flight supervision still suffers from insufficient coordination between perception, communication, identification, and control, particularly in cases involving non-cooperative targets, where tracing and verifying the identity of drone pilots is challenging. For drones that are not registered under their real names, illegally modified, or deliberately evade supervision, their flight data, location information, and operator identity are often concealed. Even when some flight target information is obtained through radio detection or radar detection, factors such as frequency hopping and encrypted transmission often make it difficult to accurately link the flight target to a specific pilot. Furthermore, drone flight activities typically rely on ground personnel for operation, and current technologies are insufficient for the coordinated use of low-altitude flight dynamic data and ground vehicle checkpoint data, making it difficult to achieve accurate tracing from flight targets to responsible parties.

[0004] Therefore, in tracing the identity of non-cooperative drone pilots, the strong concealment of regulatory targets, the difficulty in linking flight information with operator identity, and the insufficient collaborative utilization of ground spatiotemporal data have become urgent problems to be solved. Summary of the Invention

[0005] This application provides a method for tracing the identity of non-cooperative drone pilots based on vehicle spatiotemporal correlation, aiming to solve the problems of strong concealment of regulatory targets, difficulty in associating flight information with operator identity, and insufficient collaborative utilization of ground spatiotemporal data in the existing technology for tracing the identity of non-cooperative drone pilots.

[0006] This invention discloses a non-cooperative drone pilot identification method based on vehicle spatiotemporal correlation, the method comprising: Acquire feature point data of the target aircraft and form a point list with spatiotemporal attributes; wherein, the point list includes multiple target feature points; Based on the vehicle big data system, target vehicle checkpoints are filtered according to the spatial location correlation of each target feature point in the location list, and license plate number data are filtered according to the temporal correlation of each target feature point to obtain the license plate number array corresponding to each target feature point. The license plate number array is processed to obtain a target list sorted by the frequency of the license plate number across the array; Based on the target list, license plate numbers that meet the preset frequency ratio are filtered to obtain suspected license plates; Using the suspected license plate as the object of verification, a correlation analysis between the vehicle and the person was conducted to identify the suspected person; The identity of the suspected person was verified based on the facial image.

[0007] Optionally, in the above scheme, acquiring the feature point data of the target aircraft and forming a point list with spatiotemporal attributes includes: The system acquires and stores historical flight dynamics data and remote controller information data through a low-altitude surveillance system. Based on the flight dynamics history data and the remote controller information data, a feature point database is formed; The target feature points corresponding to the target aircraft are extracted from the feature point database to obtain the point list.

[0008] Optionally, in the above scheme, the flight dynamic historical data includes the latitude and longitude coordinates and takeoff time of the aircraft's takeoff point, and the latitude and longitude coordinates and landing time of the aircraft's landing point; The remote control information data includes the latitude and longitude coordinates of the remote control location and the corresponding time.

[0009] Optionally, in the above scheme, the step of extracting the target feature points corresponding to the target aircraft from the feature point database to obtain the point list includes: Based on the aircraft serial number of the target aircraft, candidate feature point data corresponding to the target aircraft are determined from the feature point database; Filter the target feature points of the target aircraft within a preset time period from the candidate feature point data; Based on the target feature points, the point list is formed.

[0010] Optionally, in the above scheme, the plurality of target feature points include the takeoff point, landing point, and remote controller point in the target aircraft's flight record.

[0011] Optionally, in the above scheme, the step of filtering target vehicle checkpoints based on the spatial location correlation of each target feature point in the point list using the vehicle big data system includes: For each target feature point, a preset range is set with the target feature point as the center, based on the planar coordinates of the target feature point. Based on the vehicle big data system, vehicle checkpoints located within the preset range are identified as target vehicle checkpoints whose spatial location is related to the target feature points.

[0012] Optionally, in the above scheme, the step of filtering license plate number data according to the time correlation of each target feature point to obtain the license plate number array corresponding to each target feature point includes: For each target feature point, a preset time period related to the time of the target feature point is determined based on the time corresponding to the target feature point. Based on the target vehicle checkpoint corresponding to the target feature point and the preset time period, extract license plate number data from the vehicle big data system; Based on the extracted license plate number data, an array of license plate numbers corresponding to the target feature points is obtained.

[0013] In the above scheme, optionally, for each target feature point, the license plate number array corresponding to the target feature point includes all license plate numbers of the target vehicle checkpoint within the preset time period that are located within the preset range.

[0014] Optionally, in the above scheme, processing the license plate number array to obtain a target list sorted by the frequency of occurrence of the license plate number across the array includes: The license plate numbers in each license plate number array are deduplicated to obtain the deduplicated license plate number array; Based on multiple deduplicated license plate number arrays, determine the frequency of each license plate number appearing across arrays; Based on the frequency of each license plate number across arrays, the license plate numbers are sorted in descending order to obtain the target list; wherein, the target list includes the license plate number, frequency of occurrence, capture time, and capture checkpoint.

[0015] Optionally, in the above scheme, the step of filtering license plate numbers that meet a preset frequency ratio based on the target list to obtain suspected license plates includes: Based on the comparison of the frequency of each license plate number appearing across arrays with the preset frequency ratio, the license plate numbers in the target list are filtered to obtain the suspected license plates.

[0016] Compared with the prior art, this application has at least the following beneficial effects: This application, based on further analysis and research into existing technical problems, recognizes that existing technologies suffer from several shortcomings in tracing the identity of pilots of non-cooperative drones. These include the strong concealment of regulatory targets, difficulty in associating flight information with operator identity, and insufficient collaborative utilization of ground spatiotemporal data. By acquiring feature point data of target aircraft and forming a list of points with spatiotemporal attributes, the previously scattered flight activity information is converged into an analytical basis capable of characterizing the key activity positions of the target aircraft. Furthermore, based on a vehicle big data system, target vehicle checkpoints are filtered according to the spatial correlation of each target feature point, and license plate number data is filtered according to the temporal correlation of each target feature point, resulting in an array of license plate numbers corresponding to each target feature point. This establishes a spatiotemporal correspondence between the aerial trajectory of the target aircraft and ground vehicle activity information. Further, by processing the license plate number array, a target list is obtained, sorted by the frequency of occurrence of license plate numbers across the array, and based on… The system filters license plate numbers that meet a preset frequency ratio from the target list to obtain suspected license plates. This allows for the identification of vehicles with a high degree of repetition and association with multiple target feature points from a large number of candidate vehicles. Based on this, the suspected license plates are used as the verification object for vehicle-person association analysis to determine suspected persons. Identity verification is then completed based on the facial images of the suspected persons, thereby linking the target aircraft, suspected vehicles, and suspected persons to form a complete identity tracing chain. Therefore, this application does not rely on directly calculating the pilot's identity or directly cracking the communication link of non-cooperative drones. Instead, it gradually narrows down the verification scope and ultimately leads to the identity of specific persons through the correlation between the feature points of the target aircraft and the spatiotemporal data of ground vehicles. This effectively solves the problems in the background technology of the strong concealment of non-cooperative drone monitoring targets, the difficulty in associating flight information with the operator's identity, and the insufficient collaborative utilization of ground spatiotemporal data. Attached Figure Description

[0017] Figure 1 A flowchart illustrating a non-cooperative drone pilot identification tracing method based on vehicle spatiotemporal correlation provided in one embodiment of this application; Figure 2 A flowchart illustrating a method for tracing the identity of non-cooperative drone pilots based on vehicle spatiotemporal big data, provided as an embodiment of this application; Figure 3 The process of forming an array of single feature points is provided in one embodiment of this application; Figure 4 The process for processing and frequency calculation of license plate number arrays provided in one embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] In one embodiment, such as Figure 1 As shown, a non-cooperative drone pilot identification method based on vehicle spatiotemporal correlation is provided, including the following steps: Acquire feature point data of the target aircraft and form a point list with spatiotemporal attributes; wherein, the point list includes multiple target feature points; Based on the vehicle big data system, target vehicle checkpoints are filtered according to the spatial location correlation of each target feature point in the location list, and license plate number data are filtered according to the temporal correlation of each target feature point to obtain the license plate number array corresponding to each target feature point. The license plate number array is processed to obtain a target list sorted by the frequency of the license plate number across the array; Based on the target list, license plate numbers that meet the preset frequency ratio are filtered to obtain suspected license plates; Using the suspected license plate as the object of verification, a correlation analysis between the vehicle and the person was conducted to identify the suspected person; The identity of the suspected person was verified based on the facial image.

[0020] This embodiment provides a non-cooperative drone pilot identification tracing method based on vehicle spatiotemporal correlation. First, characteristic point data of the target aircraft is acquired, forming a point list with spatiotemporal attributes. The point list includes multiple target characteristic points. Each target characteristic point has corresponding location and time attributes, enabling subsequent spatial and temporal correlation analysis around each target characteristic point. Target characteristic points can reflect key positions in the target aircraft's flight activities and positions related to control behaviors, thus allowing the point list to serve as the starting data for the entire tracing process.

[0021] After forming the location list, based on the vehicle big data system, target vehicle checkpoints are filtered according to the spatial location correlation of each target feature point in the list, and license plate number data is filtered according to the temporal correlation of each target feature point, resulting in a license plate number array corresponding to each target feature point. In specific implementation, for each target feature point, first, the target vehicle checkpoints spatially related to that target feature point are determined, then the license plate number data temporally related to that target feature point is determined, and the extracted license plate number data is organized into license plate number arrays according to the target feature points. Thus, each target feature point corresponds to one license plate number array, and multiple target feature points correspond to multiple license plate number arrays.

[0022] After obtaining the license plate number arrays corresponding to each target feature point, the license plate number arrays are processed to obtain a target list sorted by the frequency of occurrence of license plate numbers across arrays. During the processing, license plate numbers in multiple license plate number arrays are statistically analyzed to quantify the recurrence of different license plate numbers across multiple arrays, and the target list is formed accordingly. The target list reflects the degree of correlation between different license plate numbers and multiple target feature points, providing a basis for subsequent screening of suspected license plates.

[0023] After obtaining the target list, license plate numbers that meet a preset frequency ratio are filtered to obtain suspected license plates. Specifically, the frequency of each license plate number across multiple arrays in the target list is compared with a preset frequency ratio, and license plate numbers that meet the preset frequency ratio are retained as suspected license plates. This process further narrows down the candidate license plate numbers to those with a high degree of repetition and correlation with multiple target feature points.

[0024] After obtaining a suspected license plate, a correlation analysis is performed on the vehicle and its associated persons using the suspected license plate as the verification object to identify suspected persons. Specifically, based on images captured by the vehicle big data system, a correlation relationship is established between the vehicle and the person corresponding to the suspected license plate, and the suspected person is identified based on this relationship. This correlation relationship includes co-drivers, co-passengers, and the vehicle registrant. Thus, by focusing on the vehicle corresponding to the suspected license plate, suspected persons associated with that vehicle can be gradually identified.

[0025] After identifying a suspected individual, identity verification is performed based on their facial image. In practice, the facial image can be used to further verify the identity of suspected individuals identified through the aforementioned vehicle-person correlation analysis. This connects the target feature points, target vehicle checkpoints, license plate number arrays, and suspected license plates to the specific individual's identity. Thus, the complete tracing from the target aircraft to the suspected individual is finished.

[0026] This embodiment establishes a continuous tracing chain by progressively linking the target aircraft's feature data with the target vehicle checkpoints, license plate numbers, and captured images in the vehicle big data system. This chain extends from the target aircraft, target feature data, target vehicle checkpoints, license plate number arrays, target list, suspected license plates, to suspected individuals. This solution not only filters suspected license plates from the target list based on the frequency of their appearance across arrays, but also performs further correlation analysis between suspected license plates and individuals, and verifies the identity of suspected individuals based on their facial images. This effectively traces the identity of non-cooperative drone pilots.

[0027] In this embodiment, acquiring the feature point data of the target aircraft and forming a point list with spatiotemporal attributes includes: The system acquires and stores historical flight dynamics data and remote controller information data through a low-altitude surveillance system. Based on the flight dynamics history data and the remote controller information data, a feature point database is formed; The target feature points corresponding to the target aircraft are extracted from the feature point database to obtain the point list.

[0028] The system acquires and stores historical flight dynamics data and remote controller information data through a low-altitude surveillance system. Specifically, the low-altitude surveillance system continuously receives and saves historical flight information related to the target aircraft, as well as the remote controller information corresponding to that target aircraft. The historical flight dynamics data is used to characterize key activities of the target aircraft during flight, while the remote controller information data is used to characterize position and time information related to the target aircraft's control behavior. By acquiring and storing these two types of data, the original data foundation required for subsequent extraction of target feature points can be formed.

[0029] Based on the historical flight dynamic data and the remote controller information data, a feature point database is formed. In specific implementation, key point information from the historical flight dynamic data and remote controller point information from the remote controller information data can be organized, collected, and stored in a structured manner, and written into the feature point database according to a unified data format. Each record in the feature point database can correspond to a candidate feature point, allowing subsequent feature point retrieval around the target aircraft to be directly retrieved from a unified data source, without needing to re-parse the original monitoring data each time.

[0030] The target feature points corresponding to the target aircraft are extracted from the feature point database to obtain the point list. In specific implementation, feature points related to the target aircraft can be retrieved from the feature point database and then formed into a point list according to a unified point organization format. The formed point list is used for subsequent screening of target vehicle checkpoints and license plate number data. Each target feature point in the point list retains its spatiotemporal attributes, thus enabling the point list to support both spatial location correlation analysis and temporal correlation analysis.

[0031] This embodiment clarifies the source and formation path of the point list by introducing a low-altitude monitoring system, flight dynamic historical data, remote controller information data, and a feature point database. This ensures that the extraction of target feature points is based on structured data management, thereby providing stable, unified, and searchable data support for subsequent vehicle spatiotemporal correlation analysis.

[0032] In this embodiment, the flight dynamic history data includes the latitude and longitude coordinates and takeoff time of the aircraft's takeoff point, as well as the latitude and longitude coordinates and landing time of the aircraft's landing point; The remote control information data includes the latitude and longitude coordinates of the remote control location and the corresponding time.

[0033] The flight dynamic history data includes the latitude and longitude coordinates and takeoff time of the aircraft's takeoff point, as well as the latitude and longitude coordinates and landing time of the aircraft's landing point. Specifically, the flight dynamic history data stored in the low-altitude monitoring system should at least record the takeoff point location and takeoff time when the aircraft begins flight, and the landing point location and landing time when the aircraft ends flight. By storing the spatiotemporal information of these two key nodes—the takeoff point and landing point—the starting and ending positions of the target aircraft's flight activities can be clearly identified.

[0034] The remote controller information data includes the latitude and longitude coordinates of the remote controller location and the corresponding time. In specific implementation, the remote controller information data acquired by the low-altitude monitoring system should at least record the location of the remote controller location and the corresponding time. Since the remote controller location reflects the ground position related to the target aircraft's control behavior, including the latitude and longitude coordinates and corresponding time in the data range ensures that the subsequently generated location list simultaneously includes key flight activity locations and control-related locations.

[0035] By unifying the latitude and longitude coordinates and time corresponding to the aircraft's takeoff point, landing point, and remote controller location as the basic data for feature points, it can be ensured that each candidate feature point in the feature point database has clear spatial and temporal attributes. Thus, when extracting target feature points from the feature point database and forming a point list, each target feature point in the list can directly support spatial location correlation filtering and temporal correlation filtering.

[0036] In this embodiment, the step of extracting the target feature points corresponding to the target aircraft from the feature point database to obtain the point list includes: Based on the aircraft serial number of the target aircraft, candidate feature point data corresponding to the target aircraft are determined from the feature point database; Filter the target feature points of the target aircraft within a preset time period from the candidate feature point data; Based on the target feature points, the point list is formed.

[0037] Based on the target aircraft's serial number, candidate feature point data corresponding to the target aircraft is determined from the feature point database. Specifically, the target aircraft's serial number is input into the feature point database as a search condition, and all candidate feature point data corresponding to that serial number are retrieved from the database. The candidate feature point data refers to the set of data in the feature point database that already belongs to the target aircraft but has not yet been filtered by time criteria.

[0038] The target feature points corresponding to the target aircraft within a preset time period are selected from the candidate feature point data. Specifically, the preset time period is used as the filtering condition to perform time filtering on the aforementioned candidate feature point data, retaining only the candidate feature point data whose time falls within the preset time period, and determining the retained portion as the target feature points. Thus, the target feature points not only correspond to the target aircraft but also to the preset time period.

[0039] Based on the target feature points, a point list is formed. In specific implementation, the target feature points obtained after aircraft serial number retrieval and preset time period filtering are written into the point list in a unified format, so that each target feature point in the point list has a consistent data organization form, which facilitates subsequent point-by-point screening of target vehicles and license plate number data.

[0040] In this embodiment, the plurality of target feature points include the takeoff point, landing point, and remote controller point in the target aircraft's flight record.

[0041] The multiple target feature points include the takeoff point, landing point, and remote controller point in the target aircraft's flight record. In specific implementation, the point list formed by the multiple target feature points includes at least the takeoff point characterizing the target aircraft's takeoff position, the landing point characterizing the target aircraft's end-of-flight position, and the remote controller point characterizing the target aircraft's control-related position. That is, the multiple target feature points in the point list are not arbitrary points, but rather key points selected around the target aircraft's flight activities and control behaviors.

[0042] In practice, the takeoff point can be used to characterize the key spatiotemporal location at the start of a flight activity, the landing point can be used to characterize the key spatiotemporal location at the end of a flight activity, and the remote controller location can be used to characterize the spatiotemporal location associated with the pilot's control behavior. After incorporating these three types of locations into the location list, the location list can cover the key spatiotemporal information of the target aircraft from three aspects: the start of the flight activity, the end of the flight activity, and the ground control association.

[0043] In practical applications, the takeoff point, landing point, and remote controller location belonging to that time period can be retrieved from the feature point database based on the target aircraft's flight records within a preset time period, and then compiled into a unified target feature point list. This resulting point list can more accurately reflect the key activity trajectory of the target aircraft within the current analysis range.

[0044] In this embodiment, the step of filtering target vehicle checkpoints based on the spatial location correlation of each target feature point in the point list using the vehicle big data system includes: For each target feature point, a preset range is set with the target feature point as the center, based on the planar coordinates of the target feature point. Based on the vehicle big data system, vehicle checkpoints located within the preset range are identified as target vehicle checkpoints whose spatial location is related to the target feature points.

[0045] For each target feature point in the target feature point list, a preset range is set centered on the target feature point based on its planar coordinates. In specific implementation, for each target feature point in the point list, its planar coordinates are read, and a spatial range is set around these coordinates. The preset range can be a certain radius area centered on the target feature point, or it can be other forms of spatial retrieval area established around the target feature point.

[0046] Based on the vehicle big data system, vehicle checkpoints located within the preset range are identified as target vehicle checkpoints spatially related to the target feature point. In specific implementation, the vehicle checkpoint distribution information in the vehicle big data system is retrieved, and vehicle checkpoints falling within the preset range are identified. The retrieval results are then determined as target vehicle checkpoints spatially related to the target feature point. Based on reconnaissance experience, the range value can be set to 50 meters, thus obtaining vehicle checkpoints within a 50-meter radius centered on the target feature point.

[0047] For different target feature points, their respective preset ranges can be set, and the target vehicle checkpoints related to each target feature point in space can be obtained. Thus, each target feature point in the point list can correspond to a set of target vehicle checkpoints, providing spatial constraints for subsequent extraction of license plate number data based on time correlation.

[0048] In this embodiment, the step of filtering license plate number data according to the time correlation of each target feature point to obtain the license plate number array corresponding to each target feature point includes: For each target feature point, a preset time period related to the time of the target feature point is determined based on the time corresponding to the target feature point. Based on the target vehicle checkpoint corresponding to the target feature point and the preset time period, extract license plate number data from the vehicle big data system; Based on the extracted license plate number data, an array of license plate numbers corresponding to the target feature points is obtained.

[0049] For each target feature point, a preset time period related to the time of the target feature point is determined based on the time corresponding to the target feature point. Specifically, the time information carried by each target feature point is read, and a preset time period is set around this time information. The preset time period is used to limit the time window for subsequent extraction of license plate number data, establishing a temporal correspondence between the license plate number data and the target feature points.

[0050] Based on the target vehicle checkpoint corresponding to the target feature point and the preset time period, license plate number data is extracted from the vehicle big data system. Specifically, for each target feature point, the target vehicle checkpoint corresponding to that target feature point is invoked, and the preset time period related to the time of that target feature point is used as the time condition to extract license plate number data that meets both the spatial and temporal conditions from the vehicle big data system. In other words, the extracted license plate number data simultaneously satisfies both the conditions of "originating from the corresponding target vehicle checkpoint" and "falling within the corresponding preset time period."

[0051] Based on the extracted license plate number data, an array of license plate numbers corresponding to the target feature points is obtained. In specific implementation, all extracted license plate number data corresponding to a certain target feature point are organized into a single license plate number array. For multiple target feature points, multiple license plate number arrays are formed respectively. Thus, each target feature point can be associated with a license plate number array.

[0052] In this embodiment, for each target feature point, the license plate number array corresponding to the target feature point includes all license plate numbers of the target vehicle checkpoint within the preset time period that are located within the preset range.

[0053] For each target feature point, the corresponding license plate number array includes all license plate numbers of the target vehicle checkpoints within the preset range and within the preset time period. In specific implementation, for a given target feature point, not only are some license plate numbers that meet the conditions extracted, but all license plate numbers captured by the target vehicle checkpoints within the preset range corresponding to that target feature point and within the preset time period corresponding to that target feature point are uniformly included in the license plate number array.

[0054] If a certain feature point has a specific location coordinate, a specific time, a spatial range of 50 meters, and a time range of a preset duration before and after that time, then all license plate numbers captured by vehicle checkpoints within the preset time range corresponding to that location point, as output by the vehicle big data system, constitute a single license plate number array. Similarly, multiple license plate number arrays corresponding to multiple target feature points can be formed.

[0055] Since the license plate number array includes all license plate numbers that meet the corresponding spatial and temporal conditions, the multiple license plate number arrays corresponding to different target feature points are consistent in their formation rules, which facilitates subsequent unified deduplication, frequency statistics and sorting processing.

[0056] In this embodiment, processing the license plate number array to obtain a target list sorted by the frequency of occurrence of the license plate number across the array includes: The license plate numbers in each license plate number array are deduplicated to obtain the deduplicated license plate number array; Based on multiple deduplicated license plate number arrays, determine the frequency of each license plate number appearing across arrays; Based on the frequency of each license plate number across arrays, the license plate numbers are sorted in descending order to obtain the target list; wherein, the target list includes the license plate number, frequency of occurrence, capture time, and capture checkpoint.

[0057] The license plate numbers in each license plate number array are deduplicated to obtain a deduplicated license plate number array. In practice, each target feature point's corresponding license plate number array is processed separately, removing duplicate entries of the same license plate number within the same array, ensuring that each license plate number is retained only once in a single array. This deduplication process avoids the same vehicle being repeatedly captured at the same target vehicle checkpoint within the same preset time period, preventing duplicate counting within a single array.

[0058] Based on multiple deduplicated license plate number arrays, the frequency of each license plate number appearing across arrays is determined. Specifically, multiple deduplicated license plate number arrays are used as input. Each license plate number in the array is read sequentially, and the number of times each license plate number appears in the multiple deduplicated license plate number arrays is accumulated to obtain the frequency of each license plate number appearing across arrays. This frequency of appearance across arrays reflects how many different target feature point arrays each license plate number appears in.

[0059] Based on the frequency of each license plate number across multiple arrays, the license plate numbers are sorted in descending order to obtain the target list. The target list includes the license plate number, frequency of occurrence, capture time, and capture checkpoint. Specifically, license plate numbers with higher frequencies across multiple arrays are ranked first to form the target list, and the target list records information such as the frequency of occurrence, capture time, and capture checkpoint corresponding to each license plate number. This target list not only represents the degree of repetition and association of license plate numbers but also retains the capture context information needed for subsequent verification.

[0060] In this embodiment, the step of filtering license plate numbers that meet a preset frequency ratio based on the target list to obtain suspected license plates includes: Based on the comparison of the frequency of each license plate number appearing across arrays with the preset frequency ratio, the license plate numbers in the target list are filtered to obtain the suspected license plates.

[0061] Based on the target list, license plate numbers that meet the preset frequency ratio are filtered to obtain the suspected license plates. Specifically, for each license plate number in the target list, its cross-array occurrence frequency is read, and this cross-array occurrence frequency is compared with the preset frequency ratio. The preset frequency ratio serves as a filtering threshold to distinguish between license plate numbers with high correlation and those with low correlation.

[0062] Based on the comparison between the frequency of each license plate number appearing across multiple arrays and the preset frequency ratio, the license plate numbers in the target list are filtered to obtain the suspected license plates. Specifically, for each license plate number in the target list, if its frequency of appearance across multiple arrays meets the preset frequency ratio, the license plate number is retained and identified as a suspected license plate; if its frequency of appearance across multiple arrays does not meet the preset frequency ratio, it is not included in the suspected license plate list. This filtering process further narrows down the range of candidate vehicles from the target list.

[0063] The preset frequency weight can be defined by the user; it can be the frequency of occurrence or the proportion of the total number of feature points. For example, when there are N target feature points in the point list, any integer satisfying 1≤n≤N can be used as the frequency weight, or a certain percentage value can be used as the frequency weight. By adjusting the preset frequency weight, the strictness of the suspected license plate screening can be controlled.

[0064] This embodiment sets a preset frequency weight and compares each license plate number in the target list with the preset frequency weight. This allows multiple candidate license plate numbers to be further filtered according to their degree of repetition with multiple target feature points, thereby obtaining more targeted suspected license plates. This provides input objects for subsequent vehicle and personnel association analysis based on suspected license plates.

[0065] In one embodiment, a method for tracing the identity of non-cooperative drone pilots based on vehicle spatiotemporal big data is provided. The purpose of this embodiment is to solve the problem of difficulty in supervising and tracing non-cooperative drones. This embodiment can effectively combine drone perception technology, video recognition technology, retrieval technology and big data analysis technology to provide intelligent technical support for the verification of illegal low-altitude flights, thereby improving the effective management and control efficiency of low-altitude airspace.

[0066] To achieve the above objectives, this embodiment adopts the following technical solution: a method for tracing the identity of non-cooperative drone pilots based on vehicle spatiotemporal big data, the system comprising the following steps: S1. Establish a low-altitude flight dynamic database, store historical flight dynamic data based on the low-altitude monitoring system, store the acquired remote controller information data, and form a feature point database accordingly.

[0067] S2, retrieve and filter the characteristic points of the target aircraft to form a list of points with spatiotemporal properties. Identify the analysis object based on the aircraft serial number and filter out a list of characteristic point data that meets the requirements based on the time range.

[0068] S3, based on the search results of S2 and the vehicle big data system, filters out vehicle checkpoints that are spatially related to the feature points according to the set spatial range.

[0069] S4, based on the search results of S3 and the vehicle big data system, filters out license plate number data that are related to the vehicle checkpoint in terms of time, forming an array of license plate numbers with time and spatial information.

[0070] S5, based on the license plate number array formed by S4, processes the data in each array to form a list sorted by the frequency of occurrence of the license plate number in each array. Its fields include at least the license plate number, frequency of occurrence, capture time, and capture checkpoint.

[0071] S6, based on the data analysis of S5, sets the frequency weight and filters the license plate numbers that meet the frequency weight. The license plate numbers that meet the conditions are suspected license plates.

[0072] S7 uses suspected license plates as the verification target, performs correlation analysis between vehicles and people, identifies suspected persons, and completes identity verification based on the suspected persons' images.

[0073] The process of the method for tracing the identity of non-cooperative drone pilots based on vehicle spatiotemporal big data provided in this embodiment is as follows: Figure 2 As shown.

[0074] In step 1, a low-altitude flight dynamic database is established to save feature point data. The system relies on a low-altitude surveillance system to store historical flight dynamic data, which should at least include the takeoff and landing points of the aircraft; it also stores the acquired remote controller information data; the flight dynamic data and remote controller information data should at least include latitude and longitude coordinates and the time of information acquisition, and the spatiotemporal data of the aircraft's takeoff and landing points and the spatiotemporal data of the remote controller are the feature point data.

[0075] In step 2, the characteristic points of the target aircraft are retrieved and filtered to form a list of points with spatiotemporal properties; The analysis object is identified based on the aircraft serial number, and a list of feature points that meet the requirements is filtered based on the time range. The information in the data list includes the planar coordinates and time of three types of points in the historical flight records of the current analysis object: takeoff point, landing point, and remote controller point.

[0076] In step 3, based on the retrieval results from step 2 and the vehicle big data system, vehicle checkpoints related to the spatial location of feature points are selected according to the set range.

[0077] Specifically, based on the planar coordinates of the feature points in step 2, and considering the distribution of vehicle checkpoints in the vehicle big data system, a range value is set with the feature points as the center to filter out the associated vehicle checkpoints. The range value can be manually adjusted; for example, setting the range value to 50 meters based on the experience of investigators will filter out vehicle checkpoints within a 50-meter radius of the feature points.

[0078] In step 4, based on the vehicle checkpoints and vehicle big data system selected in step 3, license plate number data that is time-related to the vehicle checkpoints is filtered out, and an array of license plate numbers is formed according to the feature points. The time-relatedness mentioned here refers to the time period related to the feature points.

[0079] The process of forming an array from individual feature points in steps 3 and 4 is as follows: Figure 3As shown, taking a single feature point as an example, with the planar coordinates of the feature point being (x, y), the acquisition time being yyyy-mm-dd hh-mm-ss, the retrieval spatial range being set to 50 meters, and the time range being set to 30 minutes, the vehicle big data system will output all license plate numbers of all vehicle checkpoints within a 50-meter radius centered at (x, y) within 30 minutes before and after the time point yyyy-mm-dd hh-mm-ss. All the license plate numbers analyzed here are in an array.

[0080] In step 5, based on the license plate number array formed in step 4, the data in each array is deduplicated and frequency counted to form an element list sorted in descending order of the frequency of license plate numbers in each array. The list fields include at least the license plate number, frequency of occurrence, capture time, and capture checkpoint.

[0081] The method for generating a list of elements in each array whose license plate number appears in descending order of frequency is as follows: Figure 4 As shown, the algorithm first receives multiple arrays as input, then performs deduplication on each array in turn to ensure that each element is counted only once in a single array. The deduplicated elements are added to a global frequency statistics table for cumulative counting. After processing all arrays, the elements are sorted in descending order according to their frequency across arrays, and finally the sorted list of elements is output.

[0082] In step 6, based on the element list output in step 5, a frequency weight is set, and license plate numbers that meet the frequency weight are filtered. License plate numbers that meet the conditions are suspected license plates.

[0083] The frequency weight can be defined by the user. It can be the frequency of occurrence or the proportion of the total number of feature points. For example, after step 2, a total of N feature point data are output. The frequency weight can be any integer that satisfies 1≤n≤N or it can be a percentage.

[0084] In step 7, the suspected license plate is used as the verification target. Based on the images captured by the vehicle big data system, a correlation analysis between the vehicle and the person is performed to identify the suspected person. The identity of the suspected person is then verified based on their facial image. The correlation between the vehicle and the person includes the driver, passenger, and vehicle registrant.

[0085] This embodiment effectively combines UAV perception technology, video recognition technology, retrieval technology and big data analysis technology, integrates vehicle data resources such as license plate recognition and checkpoint monitoring, and associates UAV operators with suspicious vehicle information from spatial and temporal dimensions, locks down the identity of suspicious vehicles and pilots, and provides an effective solution for verifying illegal low-altitude flights and tracing the source of pilots, thereby improving the effective control efficiency of low-altitude airspace.

[0086] This embodiment innovatively combines a vehicle big data system with the needs of drone supervision. By integrating vehicle data resources such as license plate recognition and checkpoint monitoring, and through big data analysis, it links drone operators with information on suspicious vehicles, identifying suspicious vehicles and drone pilots. Compared to traditional manual patrols, this technology achieves three major breakthroughs: first, it expands regulatory capabilities at zero cost using existing infrastructure; second, it improves the accuracy of drone pilot tracing through spatiotemporal data cross-verification; and third, it forms a chain of evidence for drone pilots with multiple violations based on historical data backtracking. Furthermore, it can mine vehicle and drone pilot relationship graphs, and identify organizational networks by analyzing vehicle registration information and drivers' social relationships, providing a complete electronic evidence chain for subsequent law enforcement.

[0087] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A non-cooperative drone pilot identification method based on vehicle spatiotemporal correlation, characterized in that, The method includes: Acquire feature point data of the target aircraft and form a point list with spatiotemporal attributes; wherein, the point list includes multiple target feature points; Based on the vehicle big data system, target vehicle checkpoints are filtered according to the spatial location correlation of each target feature point in the location list, and license plate number data are filtered according to the temporal correlation of each target feature point to obtain the license plate number array corresponding to each target feature point. The license plate number array is processed to obtain a target list sorted by the frequency of the license plate number across the array; Based on the target list, license plate numbers that meet the preset frequency ratio are filtered to obtain suspected license plates; Using the suspected license plate as the object of verification, a correlation analysis between the vehicle and the person was conducted to identify the suspected person; The identity of the suspected person was verified based on the facial image.

2. The method according to claim 1, characterized in that, The process of acquiring feature point data of the target aircraft and forming a point list with spatiotemporal attributes includes: The system acquires and stores historical flight dynamics data and remote controller information data through a low-altitude surveillance system. Based on the flight dynamics history data and the remote controller information data, a feature point database is formed; The target feature points corresponding to the target aircraft are extracted from the feature point database to obtain the point list.

3. The method according to claim 2, characterized in that, The flight dynamic historical data includes the latitude and longitude coordinates and takeoff time of the aircraft's takeoff point, as well as the latitude and longitude coordinates and landing time of the aircraft's landing point; The remote control information data includes the latitude and longitude coordinates of the remote control location and the corresponding time.

4. The method according to claim 2, characterized in that, The step of extracting the target feature points corresponding to the target aircraft from the feature point database to obtain the point list includes: Based on the aircraft serial number of the target aircraft, candidate feature point data corresponding to the target aircraft are determined from the feature point database; Filter the target feature points of the target aircraft within a preset time period from the candidate feature point data; Based on the target feature points, the point list is formed.

5. The method according to claim 4, characterized in that, The multiple target feature points include the takeoff point, landing point, and remote controller point in the target aircraft's flight record.

6. The method according to claim 1, characterized in that, The method of filtering target vehicle checkpoints based on the spatial location correlation of each target feature point in the location list, using the vehicle big data system, includes: For each target feature point, a preset range is set with the target feature point as the center, based on the planar coordinates of the target feature point. Based on the vehicle big data system, vehicle checkpoints located within the preset range are identified as target vehicle checkpoints whose spatial location is related to the target feature points.

7. The method according to claim 1, characterized in that, The process of filtering license plate number data according to the time correlation of each target feature point to obtain an array of license plate numbers corresponding to each target feature point includes: For each target feature point, a preset time period related to the time of the target feature point is determined based on the time corresponding to the target feature point. Based on the target vehicle checkpoint corresponding to the target feature point and the preset time period, extract license plate number data from the vehicle big data system; Based on the extracted license plate number data, an array of license plate numbers corresponding to the target feature points is obtained.

8. The method according to claim 7, characterized in that, For each target feature point, the license plate number array corresponding to the target feature point includes all license plate numbers of the target vehicle checkpoint within the preset time period that are located within the preset range.

9. The method according to claim 1, characterized in that, The process of processing the license plate number array to obtain a target list sorted by the frequency of occurrence of the license plate number across the array includes: The license plate numbers in each license plate number array are deduplicated to obtain the deduplicated license plate number array; Based on multiple deduplicated license plate number arrays, determine the frequency of each license plate number appearing across arrays; Based on the frequency of each license plate number across arrays, the license plate numbers are sorted in descending order to obtain the target list; wherein, the target list includes the license plate number, frequency of occurrence, capture time, and capture checkpoint.

10. The method according to claim 9, characterized in that, Based on the target list, license plate numbers that meet a preset frequency ratio are filtered to obtain suspected license plates, including: Based on the comparison results of the frequency of occurrence of each license plate number across arrays with the preset frequency ratio, the license plate numbers in the target list are filtered to obtain the suspected license plates.

Images