A DBSCAN-based radar and GPS track data processing and clustering method

Abstract

The application discloses a radar and GPS track data processing and clustering method based on DBSCAN, and comprises the following steps: performing protocol analysis and structured reorganization on radar track data and GPS data, and constructing a unified space-time track matrix; realizing track reconstruction based on time continuity constraint and target number grouping; performing data integrity repair by adopting linear interpolation and outlier elimination strategy; extracting kinematic characteristics and signal characteristics of the track; performing adaptive density clustering on track points by adopting a DBSCAN algorithm based on density accessibility theory; constructing a multi-rule fusion decision model based on the clustering result, realizing interference track identification and target screening; and finally outputting a visual analysis result; and the application improves target identification accuracy and system robustness.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing and data fusion technology, and in particular to a method for processing and clustering radar and GPS track data based on DBSCAN. Background Technology

[0002] Radar, as a primary means of surveillance, boasts advantages such as all-weather operation and high precision. However, it is susceptible to noise interference in complex environments, leading to broken tracks or the generation of false targets. GPS (Global Positioning System) data offers high continuity and stability, but its update frequency and spatial accuracy are limited by satellite signals.

[0003] In practical applications, a single data source often fails to meet high reliability requirements, making multi-source data fusion an inevitable trend. However, radar and GPS data differ significantly in protocol format, spatiotemporal reference, and update frequency, making direct fusion difficult. Furthermore, due to environmental interference and sensor noise, the fused track data often exhibits uneven density distribution, broken tracks, and numerous noise points.

[0004] Traditional clustering algorithms (such as K-Means) require pre-specifying the number of clusters, making them unsuitable for scenarios with an unknown number of targets. While the classic DBSCAN algorithm does not require pre-specifying the number of clusters, it is sensitive to parameters and, when processing track data with uneven density, easily misclassifies real targets in sparse areas as noise or interference in dense areas as targets. Therefore, how to effectively integrate multi-source data and perform high-quality clustering and target recognition based on this data is a pressing problem that needs to be solved. Summary of the Invention

[0005] Purpose of the invention: The purpose of this invention is to provide a method for processing and clustering radar and GPS track data based on DBSCAN, which solves the problems of numerous noise points, track breaks, and uneven density distribution in multi-source track data.

[0006] Technical solution: The present invention provides a method for processing and clustering radar and GPS track data based on DBSCAN, comprising the following steps: Step 1: Acquire radar track data and GPS positioning data; for radar track data, extract the target's timestamp, number, spatial coordinates and signal features bit by bit by detecting the fixed frame header positioning data frame to generate structured track data; for GPS positioning data, read its time and geographical location information, and convert the GPS coordinates to a coordinate system unified with the radar data to form a track data matrix with a unified spatiotemporal reference. Step 2: Group the track data according to the target identifier. Within each group, if the time difference between adjacent track points exceeds a preset time interval threshold, or if the instantaneous velocity calculated based on adjacent points exceeds a preset velocity jump threshold, the track is determined to be interrupted and divided into different track segments. If the time interval between adjacent track points is greater than a preset scan period and less than a preset maximum allowable interpolation interval, linear interpolation is used to supplement the track points at intermediate moments to form continuous and complete track data. Step 3: Extract multidimensional features from the track data, including at least average speed, total track length, mean radar cross section, and volatility, and calculate the centroid coordinates of each track; use the DBSCAN algorithm to perform density clustering on the centroid coordinates; Step 4: Based on the clustering results, at least based on the size of the cluster to which the track belongs, the average transit density of the track in the spatial grid, the average speed of the track, and the radar cross section fluctuation rate, construct a multi-rule fusion decision model to identify and mark interfering tracks, and finally output the results.

[0007] Furthermore, in step 1, the protocol parsing of the radar track data uses the fixed frame header 0xAAAA5555 to locate the start position of the data frame, and extracts the value of the field by reading the binary value of each bit one by one, multiplying it by the corresponding bit weight, and summing the results according to the start bit and length of the field in the data word.

[0008] Furthermore, in step 1, converting the GPS coordinates to a coordinate system consistent with the radar data specifically involves using the radar station location as the origin and employing the WGS84 ellipsoid model to convert the GPS longitude, latitude, and altitude into coordinates in the East-North-Sky coordinate system centered on the radar station.

[0009] Furthermore, in step 2, the time interval threshold is 30 seconds, the speed jump threshold is 30 meters per second; the maximum allowable interpolation interval is 21 seconds, and the scan cycle is 6 seconds.

[0010] Furthermore, in step 3, the neighborhood radius of the DBSCAN algorithm is 100 meters, and the minimum number of points is 2; a point is considered a core point when its neighborhood contains no fewer than 2 points.

[0011] Furthermore, in step 4, the multi-rule fusion decision model includes: Cluster size rule: If the number of tracks contained in the cluster to which a track belongs exceeds the preset cluster size threshold, the track is determined to be interference; Grid density rule: Divide the monitoring area into several spatial grids, count the number of times each grid is passed by the flight path, calculate the average number of times each flight path passes through the grids as its average grid density, and when the density exceeds the preset density threshold, the flight path is determined to be interference; Speed ​​rule: When the average speed of a track exceeds a preset speed threshold, the track is determined to be interference; Signal fluctuation rule: When the radar cross section fluctuation rate of a track is lower than the preset fluctuation rate threshold, and the median radar cross section of the track is higher than the preset median threshold, the track is determined to be a non-natural target and marked.

[0012] Furthermore, the cluster size threshold is 2, the density threshold is 3, and the side length of the spatial grid is 30 meters.

[0013] The present invention discloses a radar and GPS track data processing and clustering system based on DBSCAN, comprising:

[0014] The data parsing module is used to acquire and parse radar track data and GPS positioning data to generate track data with a unified spatiotemporal reference.

[0015] The track reconstruction and repair module is used to segment the track based on the target identifier and time continuity, and to repair missing and abnormal data.

[0016] The feature extraction and clustering module is used to extract multidimensional features of the track and perform density clustering on the track based on the DBSCAN algorithm with weighted distance metric.

[0017] The fusion decision and output module is used to construct a multi-rule fusion decision model based on clustering results and multi-dimensional features, identify interfering flight paths, and output the screening results and visualization images.

[0018] An electronic device according to the present invention includes a processor and a memory, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements any of the methods described herein.

[0019] The present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements any of the methods described herein.

[0020] Beneficial Effects: Compared with existing technologies, this invention has the following significant advantages: This invention effectively fuses radar and GPS multi-source track data through protocol parsing, field mapping, and unified coordinate transformation, solving the problems of inconsistent data formats and coordinate systems, and improving the versatility and adaptability of data processing. Based on target number grouping and time continuity constraints, track segments are divided, and linear interpolation is used to supplement missing points, effectively restoring the continuity of broken tracks and improving the stability of feature calculation. This invention comprehensively utilizes the spatial position, velocity, and RCS features of tracks to construct a weighted scoring mechanism for interference identification and target screening. The judgment results have clear physical meaning and strong interpretability. By using DBSCAN clustering to identify spatially clustered interference track groups, combined with grid density analysis, velocity thresholds, and RCS volatility thresholds, non-target interference such as vehicles and buildings can be effectively eliminated, improving target identification accuracy. Attached Figure Description

[0021] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0022] The technical solution of the present invention will be further described below with reference to the accompanying drawings. (All algorithm experiments and verifications in the following embodiments were completed in the PyCharm integrated development environment using the Python language and the PyTorch deep learning framework. The specific parameter values ​​given in the embodiments are the preferred configurations determined by the present invention after comparative experiments and analysis on track data.)

[0023] like Figure 1 As shown, an embodiment of the present invention provides a method for processing and clustering radar and GPS track data based on DBSCAN, including the following steps:

[0024] S1. Data Analysis and Coordinate Transformation:

[0025] In this embodiment, radar track data is encapsulated in a fixed frame header format, with each frame consisting of several 32-bit unsigned integer data. The system parses the data using frame synchronization positioning and bit-by-bit field extraction. The frame header identifier H = 0xAAAA5555. The system iterates through the data stream, and when consecutive data equal to H is detected, the start position of a new frame is determined. The bit-by-bit field parsing method is as follows: Let a 32-bit data word be D, a field start bit be s, and a length be n, then the field value... This method extracts parameters such as timestamp, target number, distance, azimuth, pitch, radial velocity, and RCS, ultimately generating a structured track data structure.

[0026] GPS data parsing: Read the CSV file and extract fields such as longitude, latitude, altitude, and timestamp. Use the WGS84 ellipsoid model for coordinate transformation, converting the GPS coordinates to the ENU coordinate system with the radar station as the origin. Radar station coordinate configuration (using the Hanbi Building coordinates as an example in this embodiment): radarPos.latitude (radar station latitude) = 32.041850, radarPos.longitude (radar station longitude) = 118.716537, radarPos.height (radar station altitude) = 80.

[0027] S2. Track Reconstruction and Continuity Determination:

[0028] Because radar track data may exhibit number reuse or time discontinuity, this invention groups track points and determines time continuity to construct physically continuous track segments. The specific steps are: grouping track points by target ID; sorting each group of tracks by time; and calculating the time difference between adjacent points. ;like (Time break threshold, in this embodiment) If the time exceeds 30 seconds, the track is considered broken and divided into new track segments. At the same time, spatial distance and velocity are calculated. If the calculated velocity exceeds MAX_PHYSICAL_SPEED_JUMP (maximum physical velocity jump threshold, set to 30m / s in this embodiment), track segmentation is also performed.

[0029] S3, Track Interpolation and Point Supplementation:

[0030] When short time intervals are missing between adjacent waypoints, interpolation is used to restore data continuity and improve the stability of feature calculation. The interpolation condition is: when Linear interpolation is performed at that time. For the scan cycle, This embodiment uses the maximum interpolation time window. =6.0s, =21s). The interpolation formula is: The parameters such as distance, bearing, velocity, and RCS are supplemented by additional data points. Let be the time to be interpolated, and 'a' be the interpolation ratio coefficient, representing the relative position of the time to be interpolated between the two points in the time interval.

[0031] S4. Track Feature Extraction:

[0032] For each track, calculate the following characteristics:

[0033] (1) Kinematic characteristics: average velocity ,in Instantaneous velocity at each point; total length of the track. ,in , These are coordinates based on the ENU coordinate system.

[0034] (2) Signal characteristics: mean RCS RCS volatility ,in The standard deviation of the RCS is given.

[0035] (3) Spatial characteristics: track centroid It is used for cluster analysis.

[0036] S5. Spatial grid density statistics:

[0037] The monitoring area is divided into grids (grid resolution in this embodiment) =30 meters), count the number of times each grid is traversed by the track. The average density of the i-th track is calculated using the formula: ,in For the grid in coordinates The number of visits at that location. When ( This embodiment uses the average density threshold. When the value is 3, it is marked as a high-density track to identify regional interference.

[0038] S6, DBSCAN cluster analysis:

[0039] Density clustering is performed on the centroids of the tracks to identify spatially clustered interfering track groups. The core decision criterion is: let the neighborhood radius ε (in this embodiment) =100 meters), minimum number of points MinPts (in this embodiment) =2), when the number of points in the neighborhood of a point p satisfies At that time, this point is the core point. If the clustered track cluster contains more than a threshold... (Cluster size threshold, in this embodiment) If there are 2 or more tracks, it is considered an interference track cluster.

[0040] S7. Multi-rule fusion decision:

[0041] The following rules are used to make the final classification decision on the tracks:

[0042] (71) Cluster determination: When a track belongs to a cluster containing more than the cluster size threshold, it is determined to be interference;

[0043] (72) Grid density determination: When the average density of the track exceeds the density threshold, it is determined to be interference;

[0044] (73) Speed ​​determination: When the average speed of the track exceeds the speed threshold, it is determined to be interference;

[0045] (74) RCS volatility determination: When RCS volatility is extremely low ( And the RCS value is relatively large (RCS median) When a path is identified as a non-natural target, it is determined to be a non-natural target. The final output result is Result = {is_clutter,reason}, achieving interpretable path selection. The specific values ​​of the parameters in this invention were obtained through simulation experiments.

[0046] The physical meanings of the parameters are shown in Tables 1 and 2. Table 1 Physical meaning of parameters 1 ; Table 2 Physical Meaning of Parameters .

Claims

1. A method for processing and clustering radar and GPS track data based on DBSCAN, characterized in that, Includes the following steps: Step 1: Acquire radar track data and GPS positioning data; for radar track data, extract the target's timestamp, number, spatial coordinates, and signal characteristics bit by bit by detecting the fixed frame header positioning data frame to generate structured track data; for GPS positioning data, read its time and geographical location information, and convert the GPS coordinates to a coordinate system unified with the radar data to form a track data matrix with a unified spatiotemporal reference; for radar track data protocol parsing, locate the starting position of the data frame with the fixed frame header 0xAAAA5555, and extract the field values ​​by reading the binary value of each bit one by one and multiplying it by the corresponding bit weight and then summing it according to the starting position and length of the field in the data word; converting the GPS coordinates to a coordinate system unified with the radar data is specifically done by using the radar station location as the origin and using the WGS84 ellipsoid model to convert the GPS longitude, latitude, and altitude into coordinates in the East-North-Sky coordinate system centered on the radar station; Step 2: Group the track data according to the target identifier. Within each group, if the time difference between adjacent track points exceeds a preset time interval threshold, or if the instantaneous velocity calculated based on adjacent points exceeds a preset velocity jump threshold, the track is determined to be interrupted and divided into different track segments. If the time interval between adjacent track points is greater than a preset scan period and less than a preset maximum allowable interpolation interval, linear interpolation is used to supplement the track points at intermediate moments to form continuous and complete track data. Step 3: Extract multidimensional features from the track data, including at least average speed, total track length, mean radar cross section, and volatility, and calculate the centroid coordinates of each track; use the DBSCAN algorithm to perform density clustering on the centroid coordinates; the neighborhood radius of the DBSCAN algorithm is 100 meters, and the minimum number of points is 2; a point is considered a core point when its neighborhood contains no less than 2 points. Step 4: Based on the clustering results, at least based on the size of the cluster to which the track belongs, the average transit density of the track in the spatial grid, the average speed of the track, and the radar cross section fluctuation rate, construct a multi-rule fusion decision model to identify and mark interfering tracks, and finally output the filtered analysis results. Multi-rule fusion decision models include: Cluster size rule: If the number of tracks contained in the cluster to which a track belongs exceeds the preset cluster size threshold, it is judged as interference; Grid density rule: Divide the monitoring area into several spatial grids, count the number of times each grid is passed by the flight path, and calculate the average number of times each flight path passes through the grids as its average grid density. When the density exceeds the preset density threshold, it is judged as interference. Speed ​​rule: When the average speed of a flight path exceeds a preset speed threshold, it is considered interference; Signal fluctuation rule: When the radar cross section fluctuation rate of a flight path is lower than a preset fluctuation rate threshold, and the median radar cross section of the flight path is higher than a preset median threshold, it is determined to be a non-natural target and marked; where, Track feature extraction: Calculate the following features for each track: Kinematic characteristics: average velocity ,in Instantaneous velocity at each point; total length of the track. ,in , Coordinates are based on the ENU coordinate system; Signal characteristics: mean RCS RCS volatility ,in RCS standard deviation; Spatial characteristics: track centroid This is used for cluster analysis of spatial grid density statistics: the monitoring area is divided into grids, and the number of times each grid is traversed by a flight track is counted; the average density of the i-th flight track is calculated using the following formula: ,in For the grid in coordinates Access count at the location; when At that time, among them, The average density threshold is used to mark high-density tracks for identifying regional interference. DBSCAN clustering analysis: Density clustering of track centroids to identify spatially clustered interfering track clumps; the core decision criterion is: Let the neighborhood radius be ε, and the minimum number of points be MinPts. When the number of points in the neighborhood of a point p satisfies... At that time, the core point is the location of the clustered track clusters; if the clustered track clusters contain more than a threshold value... If a single track is identified, it is classified as an interfering track cluster; the final classification decision is made based on the following rules: Clustering determination: When a track belongs to a cluster containing more than a cluster size threshold, it is determined to be interference; Grid density determination: When the average density of a track exceeds a density threshold, it is determined to be interference; Velocity determination: When the average velocity of a track exceeds a velocity threshold, it is determined to be interference; RCS fluctuation determination: When... and If the target is not natural, it is judged as a non-natural target; the final output is the judgment result Result = {is_clutter, reason}.

2. The method for processing and clustering radar and GPS track data based on DBSCAN according to claim 1, characterized in that, In step 2, the time interval threshold is 30 seconds, the speed jump threshold is 30 meters per second, the maximum allowable interpolation interval is 21 seconds, and the scan cycle is 6 seconds.

3. The method for processing and clustering radar and GPS track data based on DBSCAN according to claim 1, characterized in that, The cluster size threshold is 2, the density threshold is 3, and the side length of the spatial grid is 30 meters.

4. A radar and GPS track data processing and clustering system based on DBSCAN, implemented using the method described in any one of claims 1-3, characterized in that, include: The data parsing module is used to acquire and parse radar track data and GPS positioning data to generate track data with a unified spatiotemporal reference. The track reconstruction and repair module is used to segment the track based on the target identifier and time continuity, and to repair missing and abnormal data. The feature extraction and clustering module is used to extract multidimensional features of the track and perform density clustering on the track based on the DBSCAN algorithm with weighted distance metric. The fusion decision and output module is used to construct a multi-rule fusion decision model based on clustering results and multi-dimensional features, identify interfering flight paths, and output the screening results and visualization images.

Images