Analysis method for generating radar detection power based on target track big data

Abstract

The invention discloses an analysis method for generating radar detection power based on target track big data, and relates to the technical field of radar detection. According to the method, firstly, comprehensive and accurate evaluation of radar detection performance is realized by using historical target track big data and radar original detection trace point data, and the problem of deviation between a traditional theoretical model and an actual environment is solved; secondly, in combination with a digital elevation model and earth surface coverage type data, complex environmental factors such as terrain shielding and a multi-path effect are considered, and the reliability and applicability of an evaluation result are improved; furthermore, a binomial distribution statistical model and a Bayesian estimation method are adopted, so that the calculation accuracy of the detection probability and the measurement precision is improved; a visual display function is realized, and a generated radar detection power coverage map and a generated precision distribution map provide visual support for optimized deployment of a radar system; through model verification and iteration steps based on K-S test, continuous optimization and adaptability of the model are ensured.

Description

Technical Field

[0001] This invention relates to the field of radar detection technology, specifically to an analysis method for generating radar detection power based on big data of target tracks. Background Technology

[0002] In the field of radar detection technology, accurately assessing the detection capability of radar under different environmental conditions is crucial. Traditional radar detection performance evaluation methods often rely on theoretical models and laboratory tests. While these methods can provide some reference, in practical applications, due to the complex and ever-changing battlefield environment, terrain, weather conditions, and electromagnetic interference, there are often significant deviations between theoretical models and actual conditions.

[0003] The shortcomings of existing technology: 1. Deviation between theoretical models and actual environment: Traditional radar detection performance evaluation mainly relies on theoretical models and simulation experiments. These models are usually constructed with a large number of simplified assumptions, making it difficult to fully consider the complex factors in the actual environment, such as terrain masking, multipath effects, atmospheric attenuation, etc., resulting in a large deviation between the evaluation results and the actual situation.

[0004] 2. Limitations of measured data: Although measured data can more realistically reflect the detection performance of radar, traditional measured data collection methods are often limited by factors such as test scenarios, target types and quantities, making it difficult to cover all possible detection conditions and target states, resulting in incomplete data and a lack of universality in the evaluation results.

[0005] 3. Complexity of Data Processing and Analysis: Radar measurement data typically contains a large amount of noise and interference, requiring complex data processing and analysis to extract useful information. Traditional methods often struggle to balance computational efficiency and accuracy when processing this data, resulting in time-consuming evaluation processes and inaccurate results.

[0006] 4. Poor adaptability to dynamic environments: The radar detection environment is dynamically changing. Changes in weather conditions and target motion can affect radar detection performance. Traditional evaluation methods often struggle to adapt to these changes in real time, resulting in evaluation results that lag behind the actual situation. Summary of the Invention

[0007] To address the aforementioned technical problems, this technical solution provides an analysis method for generating radar detection capabilities based on target trajectory big data. The solution resolves the issues raised in the background section.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect of the invention, a method for analyzing radar detection capability based on target trajectory big data is provided, comprising: Acquire a large dataset of historical target tracks within the target area. The dataset contains high-precision position coordinate sequences, velocity vector sequences, heading angle sequences, and millisecond-level timestamps of real targets. Simultaneously, acquire the original detection point data of the radar system in the same time period for the area. The point data includes the amplitude, signal-to-noise ratio (SNR), range, azimuth, and pitch measurements of the original video echo. Align the real target track data with the original radar detection point data in the time dimension, calculate the three-dimensional Euclidean distance between each real target position and the radar detection point in the spatial dimension based on the nearest neighbor algorithm, and establish track-point association pairs. Set a distance threshold and a velocity threshold. When the Euclidean distance is less than the set threshold and the velocity vector angle is less than the set angle, it is determined that the real target position has been successfully detected by the radar and marked as a "successful detection event". Otherwise, it is marked as a "missed event". Based on the location coordinates of all real targets, combined with digital elevation model (DEM) data and land cover type data, the visibility between the radar and the target, the obstruction angle, and the multipath reflection coefficient are calculated. Based on successful detection events, missed detection events, line-of-sight results, radar system nominal parameters, and ambient noise floor, a binomial distribution statistical model is used to calculate the radar detection probability and measurement accuracy values ​​at different azimuth angles, elevation angles, and range thresholds. The calculated detection probability and accuracy values ​​are mapped to a geographic coordinate system and corrected using an atmospheric attenuation model to generate a radar detection coverage map and an accuracy distribution map.

[0009] Preferably, after obtaining the large dataset of historical target tracks within the target area, the following data cleaning and coordinate system transformation steps are performed: The real target trajectory data is smoothed using a Kalman filter based on a uniform motion model. The process noise covariance matrix Q is set to a diagonal matrix, and the following covariance prediction formula is executed: ; in, The prior estimate is the error covariance matrix; This is the state transition matrix; The posterior estimation error covariance matrix; This is the process noise covariance matrix, whose diagonal elements are the variances of the position components. , , and velocity component variance , , ; For the raw radar detection point data, the instantaneous signal-to-noise ratio (SNR) is first calculated, and then points with an SNR less than 13 dB are removed. The SNR calculation formula is as follows: ; in, This refers to the peak power of the video pulse. This refers to the receiver's thermal noise power. The cleaned real target WGS-84 latitude and longitude coordinates (B, L, H) are transformed with the radar station center coordinate system (0, 0, 0) using the following coordinate transformation formula: ; in,( () represents the coordinates in the northeast-northeast coordinate system; This is a 3x3 rotation matrix calculated from the latitude and longitude of the radar station; The semi-major axis of the Earth is taken as 6,378,137 meters. Earth's eccentricity; ( (B, L, H) represents the coordinates of the radar station in the geocentric coordinate system; (B, L, H) represents the latitude, longitude, and elevation of the target point.

[0010] Preferably, establishing the track-point association pair specifically includes: With radar scanning cycle Based on this, a time tolerance window is constructed. ; If a real target track point exists within the window, the radar scan time is calculated using cubic spline interpolation. The corresponding precise location of the target ( ); Calculate the spherical coordinate distance difference between the interpolated position and the radar mark. , , ; Construct the association cost function C and perform the following calculation: ; in, This represents the distance difference. This is the azimuth difference value; This is the pitch difference; For radar range resolution; This refers to the radar azimuth resolution. For radar elevation resolution; The weighting coefficients are and satisfy the following conditions: ; When the cost function C is less than the preset threshold of 1.5, the association is confirmed to be successful; otherwise, it is considered as unassociated.

[0011] Preferably, the setting of the distance threshold and determination of the detection status further includes logic for eliminating "undetectable areas": Construct a radar resolution cell grid with a grid size of 500 meters in the range direction and 1 degree in the azimuth direction; For each real target location, calculate its corresponding radar beam center pointing angle; If the visibility calculation result of the actual target location is that it is blocked, that is, the line connecting the radar and the target intersects with the terrain elevation surface, it is directly marked as an "undetectable event" and is not included in the denominator of the detection probability statistics. If the actual target location is in a line-of-sight state but no matching point is found, and there are no clutter points within 5 resolution cells around the location, it is marked as a "missed event". If the actual target location is in a line-of-sight state but no matching point is found, and there are high-intensity clutter points around it (i.e., the clutter power exceeds the noise floor by 20dB), it is marked as a "masking missed event" and is counted separately but not included in the radar's inherent detection performance. The formula for calculating the difference between clutter power and noise floor is as follows: ; in, clutter power; The noise floor power; the judgment condition is... >20dB.

[0012] Preferably, the calculation of the line-of-sight and obstruction angle between the radar and the target specifically includes: On the line connecting the radar station and the target point, N terrain profile points are sampled at equal intervals of no more than 100 meters. Ground elevation of each sampling point was extracted from DEM data using bilinear interpolation. ; Calculate the radar line of sight at the sampling point The theoretical height at the point Execute the following formula: ; in, This refers to the elevation of the radar antenna. Target elevation; The slant range of the sampling point from the radar; This represents the total slant range between the radar and the target. Calculate the radius of the first Fresnel zone of the radar beam Execute the following formula: ; in, The operating wavelength of the radar; If there are arbitrary sampling points satisfy > If so, the target is determined to be in the radar line-of-sight blind zone, and the shielding angle is calculated. : ; in, It is the shading angle.

[0013] Preferably, the step of using a binomial distribution statistical model to calculate the radar detection probability value at different azimuth angles, elevation angles, and range thresholds specifically includes: The target airspace is divided into M azimuth sectors and N range gate units to form an M×N two-dimensional statistical matrix, where the azimuth resolution is 1 degree and the range gate width is the distance corresponding to the radar pulse width. For each two-dimensional grid cell Count the total number of real targets that fall into this unit. ; Count the number of targets that fell into this cell and were marked as "detection successful events". ; Calculate the detection probability of this cell. : ; for For small sample units, the probability values ​​are corrected using Bayesian estimation, and the following formula is applied: ; in, =1, =1 is a priori parameter; Simultaneously calculate the confidence interval for this probability value. , The confidence level is set at 95%, and the calculation formula is as follows: ; in, The quantiles of the standard normal distribution are given at a 95% confidence level. =1.96.

[0014] Preferably, the specific analysis steps for radar measurement accuracy are as follows: Extract the actual target location coordinates from the "successful detection event". radar reported point coordinates ; Transform the coordinates to the radar station's northeast-sky coordinate system; Calculate range error : ; in, This refers to the range error; Calculate azimuth error : ; in, This refers to the azimuth error; Calculate the standard deviation of the range vector error within each range-azimuth grid. and the standard deviation of azimuth error : ; ; Where N is the number of samples in the grid; and These are the mean values ​​of distance and azimuth errors, respectively. And and This serves as the precision boundary value for the grid.

[0015] Preferably, the generation of the radar detection coverage map includes the following visualization and data post-processing steps: Obtain the detection probability and accuracy values ​​of all calculated grid cells; The Gaussian kernel density estimation method is used to smooth the discrete grid data, and the kernel function bandwidth h is set to 2 grid cell widths. The probability contour levels were set to 0.1, 0.5, and 0.9, and the closing polygon boundaries corresponding to the probability values ​​were extracted using the Marching Squares algorithm. Closed boundaries are overlaid onto the electronic map layer, and map tile data is stored using a quadtree structure. Color values ​​are assigned based on probability values. The color mapping table uses the HSV color space, where the H component gradually changes from 0 degrees to 120 degrees, the S component is fixed at 100%, and the V component is negatively correlated with the precision value; the lower the precision, the lower the brightness.

[0016] Preferably, after generating the radar detection coverage map, the method further includes model verification and iteration steps based on the KS test: Target trajectory data from the validation period that was not involved in modeling were selected as the test set, with a sample size of no less than 20% of the total data volume. Input the real target locations from the test set into the generated power coverage map, and read the theoretical detection probability for the corresponding locations. ; Set detection threshold Construct a confusion matrix if And if the radar actually detects a target, it is recorded as a true positive (TP); if However, the radar failed to detect the target, which was recorded as a false negative (FN). And if the radar does not detect the target, it is recorded as a true negative TN; if However, the radar detected the target and recorded it as a false positive (FP). Calculate the KS statistic ,in To detect the probability distribution function of successful samples, Let be the probability distribution function of the missed samples; If the KS statistic If the value is greater than 0.2, the model distribution difference is considered significant, and the distance threshold parameter needs to be adjusted and the steps repeated until... The value is less than or equal to 0.2.

[0017] In a second aspect of the invention, an analysis system for generating radar detection power based on target trajectory big data is also provided, comprising: The acquisition module is used to acquire a large dataset of historical target tracks within the target area. The dataset includes a high-precision sequence of position coordinates, velocity vector, heading angle, and millisecond-level timestamps of real targets. At the same time, it acquires the original detection point data of the radar system in the area within the same time period. The point data includes the amplitude, signal-to-noise ratio (SNR), range, azimuth, and pitch measurements of the original video echo. The module is used to align real target track data with radar original detection point data in the time dimension, calculate the three-dimensional Euclidean distance between each real target position and radar detection point in the spatial dimension based on the nearest neighbor algorithm, and establish track-point association pairs. The determination module is used to set a distance threshold and a velocity threshold. When the Euclidean distance is less than the set threshold and the velocity vector angle is less than the set angle, the true target position is determined to be successfully detected by the radar and marked as a "detection success event". Otherwise, it is marked as a "missed event". The first calculation module is used to calculate the line-of-sight between the radar and the target, the obstruction angle, and the multipath reflection coefficient based on the location coordinates of all real targets, combined with digital elevation model (DEM) data and land cover type data. The second calculation module is used to calculate the detection probability value and measurement accuracy value of the radar under different azimuth angles, elevation angles and range gates based on successful detection events, missed detection events, line-of-sight results, radar system nominal parameters and environmental noise floor, using a binomial distribution statistical model. The generation module maps the calculated detection probability value and accuracy value to a geographic coordinate system, and corrects them by combining an atmospheric attenuation model to generate a radar detection power coverage map and an accuracy distribution map.

[0018] Compared with existing technologies, this invention provides an analysis method for generating radar detection power based on target trajectory big data, which has the following beneficial effects: This invention firstly achieves a comprehensive and accurate assessment of radar detection performance by utilizing historical target trajectory big data and raw radar detection point data, overcoming the problem of discrepancies between traditional theoretical models and actual environments. Secondly, by combining digital elevation models and land cover type data, it considers complex environmental factors such as terrain masking and multipath effects, improving the reliability and applicability of the assessment results. Furthermore, by employing a binomial distribution statistical model and Bayesian estimation, high-precision modeling is achieved, improving the accuracy of detection probability and measurement precision calculations. In addition, this method also features visualization capabilities; the generated radar detection power coverage map and precision distribution map provide intuitive support for the optimized deployment of radar systems. Finally, through model verification and iteration steps based on the KS test, the continuous optimization and adaptability of the model are ensured, further enhancing its practical value. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the method flow of S101-S106 in this invention; Figure 2 This is a schematic diagram of the method flow for S201-S203 in this invention; Figure 3 This is a schematic diagram of the method flow for S301-S305 in this invention; Figure 4 This is a schematic diagram of the method flow for S401-S405 in this invention; Figure 5 This is a schematic diagram of the method flow for S501-S505 in this invention. Detailed Implementation

[0020] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0021] Example 1 Please refer to Figure 1 As shown, in a first aspect of the present invention, a method for analyzing radar detection power based on target trajectory big data is provided, comprising: S101. Acquire a large dataset of historical target tracks within the target area. The dataset contains high-precision position coordinate sequences, velocity vector sequences, heading angle sequences, and millisecond-level timestamps of real targets. At the same time, acquire the original detection point data of the radar system in the area within the same time period. The point data includes the amplitude, signal-to-noise ratio (SNR), range, azimuth, and pitch measurements of the original video echo. S102. Align the real target track data with the original radar detection point data in the time dimension, calculate the three-dimensional Euclidean distance between each real target position and the radar detection point in the spatial dimension based on the nearest neighbor algorithm, and establish track-point association pairs. S103. Set a distance threshold and a velocity threshold. When the Euclidean distance is less than the set threshold and the velocity vector angle is less than the set angle, it is determined that the real target position has been successfully detected by the radar and marked as a "successful detection event". Otherwise, it is marked as a "missed event". S104. Based on the location coordinates of all real targets, combined with digital elevation model (DEM) data and land cover type data, calculate the line-of-sight between the radar and the target, the obstruction angle, and the multipath reflection coefficient. S105. Based on successful detection events, missed detection events, line-of-sight results, radar system nominal parameters, and ambient noise floor, a binomial distribution statistical model is used to calculate the radar's detection probability and measurement accuracy values ​​at different azimuth angles, elevation angles, and range thresholds. S106. Map the calculated detection probability value and accuracy value to the geographic coordinate system, and correct them in combination with the atmospheric attenuation model to generate a radar detection power coverage map and accuracy distribution map.

[0022] As will be understood by those skilled in the art, this invention firstly achieves a comprehensive and accurate assessment of radar detection performance by utilizing historical target trajectory big data and raw radar detection point data, overcoming the problem of discrepancies between traditional theoretical models and actual environments. Secondly, by combining digital elevation models and land cover type data, it considers complex environmental factors such as terrain masking and multipath effects, improving the reliability and applicability of the assessment results. Furthermore, by employing a binomial distribution statistical model and Bayesian estimation, high-precision modeling is achieved, improving the accuracy of detection probability and measurement precision calculations. In addition, this method also has a visualization function; the generated radar detection power coverage map and precision distribution map provide intuitive support for the optimized deployment of radar systems. Finally, through model verification and iteration steps based on the KS test, the continuous optimization and adaptability of the model are ensured, further enhancing its practical value.

[0023] Please refer to Figure 2 As shown, after obtaining the large dataset of historical target tracks within the target area, the following data cleaning and coordinate system transformation steps are performed: S201. Smooth the real target trajectory data using Kalman filtering based on a uniform motion model, and set the process noise covariance matrix Q as a diagonal matrix, then execute the following covariance prediction formula: ; in, The prior estimate is the error covariance matrix; This is the state transition matrix; The posterior estimation error covariance matrix; This is the process noise covariance matrix, whose diagonal elements are the variances of the position components. , , and velocity component variance , , ; S202. For the original radar detection point data, first calculate the instantaneous signal-to-noise ratio (SNR), then remove points with an SNR less than 13dB. The SNR calculation formula is as follows: ; in, This refers to the peak power of the video pulse. This refers to the receiver's thermal noise power. S203. Convert the cleaned real target WGS-84 latitude and longitude coordinates (B, L, H) to the radar station center coordinate system (0, 0, 0) by performing the following coordinate transformation formula: ; in,( () represents the coordinates in the northeast-northeast coordinate system; This is a 3x3 rotation matrix calculated from the latitude and longitude of the radar station; The semi-major axis of the Earth is taken as 6,378,137 meters. Earth's eccentricity; ( (B, L, H) represents the coordinates of the radar station in the geocentric coordinate system; (B, L, H) represents the latitude, longitude, and elevation of the target point.

[0024] Please refer to Figure 3 As shown, establishing track-point association pairs specifically includes: S301, with radar scanning cycle Based on this, a time tolerance window is constructed. ; S302. If a real target track point exists within the window, the radar scan time is calculated using cubic spline interpolation. The corresponding precise location of the target ( ); S303. Calculate the spherical coordinate distance difference between the interpolated position and the radar mark. , , ; S304. Construct the correlation cost function C and perform the following calculation: ; in, This represents the distance difference. This is the azimuth difference value; This is the pitch difference; For radar range resolution; This refers to the radar azimuth resolution. For radar elevation resolution; The weighting coefficients are and satisfy the following conditions: ; S305. When the cost function C is less than the preset threshold of 1.5, the association is confirmed to be successful; otherwise, it is considered as no association.

[0025] Please refer to Figure 4 As shown, the system sets a distance threshold and determines the detection status, and also includes logic for removing "undetectable areas": S401. Construct a radar resolution cell grid, with the grid size set to 500 meters in the range direction and 1 degree in the azimuth direction; S402. For each real target location, calculate its corresponding radar beam center pointing angle; S403. If the visibility calculation result of the actual target location is that it is blocked, that is, the line connecting the radar and the target intersects with the terrain elevation surface, it is directly marked as an "undetectable event" and is not included in the denominator of the detection probability statistics. S404. If the actual target location is in a line-of-sight state but no matching point is found, and there are no clutter points within 5 resolution cells around the location, it is marked as a "missed event". S405. If the actual target location is in a line-of-sight state but no matching point is found, and there are high-intensity clutter points around it, i.e., the clutter power exceeds the noise floor by 20dB, then it is marked as a "masking missed event", which is counted separately but not included in the radar's inherent detection performance. The formula for calculating the difference between clutter power and noise floor is as follows: ; in, clutter power; The noise floor power; the judgment condition is... >20dB.

[0026] Please refer to Figure 5 As shown, the calculation of the line-of-sight and obstruction angle between the radar and the target specifically includes: S501. On the line connecting the radar station and the target point, sample N terrain profile points at equal intervals of no more than 100 meters. S502. Extract the ground elevation of each sampling point from the DEM data using bilinear interpolation. ; S503, Calculate the radar line of sight at the sampling point The theoretical height at the point Execute the following formula: ; in, This refers to the elevation of the radar antenna. Target elevation; The slant range of the sampling point from the radar; This represents the total slant range between the radar and the target. S504, Calculate the radius of the first Fresnel zone of the radar beam. Execute the following formula: ; in, The operating wavelength of the radar; S505, If any sampling point exists satisfy > If so, the target is determined to be in the radar line-of-sight blind zone, and the shielding angle is calculated. : ; in, It is the shading angle.

[0027] The detection probability values ​​of the radar at different azimuth, elevation, and range angles are calculated using a binomial distribution statistical model, specifically including: The target airspace is divided into M azimuth sectors and N range gate units to form an M×N two-dimensional statistical matrix, where the azimuth resolution is 1 degree and the range gate width is the distance corresponding to the radar pulse width. For each two-dimensional grid cell Count the total number of real targets that fall into this unit. ; Count the number of targets that fell into this cell and were marked as "detection successful events". ; Calculate the detection probability of this cell. : ; for For small sample units, the probability values ​​are corrected using Bayesian estimation, and the following formula is applied: ; in, =1, =1 is a priori parameter; Simultaneously calculate the confidence interval for this probability value. , The confidence level is set at 95%, and the calculation formula is as follows: ; in, The quantiles of the standard normal distribution are given at a 95% confidence level. =1.96.

[0028] Specific steps for analyzing radar measurement accuracy: Extract the actual target location coordinates from the "successful detection event". radar reported point coordinates ; Transform the coordinates to the radar station's northeast-sky coordinate system; Calculate range error : ; in, This refers to the range error; Calculate azimuth error : ; in, This refers to the azimuth error; Calculate the standard deviation of the range vector error within each range-azimuth grid. and the standard deviation of azimuth error : ; ; Where N is the number of samples in the grid; and These are the mean values ​​of distance and azimuth errors, respectively. And and This serves as the precision boundary value for the grid.

[0029] Generating a radar detection coverage map includes the following visualization and data post-processing steps: Obtain the detection probability and accuracy values ​​of all calculated grid cells; The Gaussian kernel density estimation method is used to smooth the discrete grid data, and the kernel function bandwidth h is set to 2 grid cell widths. The probability contour levels were set to 0.1, 0.5, and 0.9, and the closing polygon boundaries corresponding to the probability values ​​were extracted using the Marching Squares algorithm. Closed boundaries are overlaid onto the electronic map layer, and map tile data is stored using a quadtree structure. Color values ​​are assigned based on probability values. The color mapping table uses the HSV color space, where the H component gradually changes from 0 degrees to 120 degrees, the S component is fixed at 100%, and the V component is negatively correlated with the precision value; the lower the precision, the lower the brightness.

[0030] After generating the radar detection coverage map, the process also includes model verification and iteration steps based on the KS test: Target trajectory data from the validation period that was not involved in modeling were selected as the test set, with a sample size of no less than 20% of the total data volume. Input the real target locations from the test set into the generated power coverage map, and read the theoretical detection probability for the corresponding locations. ; Set detection threshold Construct a confusion matrix if And if the radar actually detects a target, it is recorded as a true positive (TP); if However, the radar failed to detect the target, which was recorded as a false negative (FN). And if the radar does not detect the target, it is recorded as a true negative TN; if However, the radar detected the target and recorded it as a false positive (FP). Calculate the KS statistic ,in To detect the probability distribution function of successful samples, Let be the probability distribution function of the missed samples; If the KS statistic If the value is greater than 0.2, the model distribution difference is considered significant, and the distance threshold parameter needs to be adjusted and the steps repeated until... The value is less than or equal to 0.2.

[0031] In a second aspect of the invention, an analysis system for generating radar detection power based on target trajectory big data is also provided, comprising: The acquisition module is used to acquire a large dataset of historical target tracks within the target area. The dataset contains high-precision position coordinate sequences, velocity vector sequences, heading angle sequences, and millisecond-level timestamps of real targets. At the same time, it acquires the original detection point data of the radar system in the area within the same time period. The point data includes the amplitude, signal-to-noise ratio (SNR), range, azimuth, and pitch measurements of the original video echo. The module is used to align real target track data with radar original detection point data in the time dimension, calculate the three-dimensional Euclidean distance between each real target position and radar detection point in the spatial dimension based on the nearest neighbor algorithm, and establish track-point association pairs. The determination module is used to set distance thresholds and velocity thresholds. When the Euclidean distance is less than the set threshold and the velocity vector angle is less than the set angle, the true target position is determined to be successfully detected by the radar and marked as a "detection success event". Otherwise, it is marked as a "missed event". The first calculation module is used to calculate the line of sight between the radar and the target, the obstruction angle, and the multipath reflection coefficient based on the location coordinates of all real targets, combined with digital elevation model (DEM) data and land cover type data. The second calculation module is used to calculate the detection probability value and measurement accuracy value of the radar under different azimuth angles, elevation angles and range gates based on successful detection events, missed detection events, line-of-sight results, radar system nominal parameters and environmental noise floor, using a binomial distribution statistical model. The generation module maps the calculated detection probability and accuracy values ​​to a geographic coordinate system and corrects them using an atmospheric attenuation model to generate a radar detection coverage map and an accuracy distribution map.

[0032] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for analyzing radar detection capability based on target trajectory big data, characterized in that, include: Acquire a large dataset of historical target tracks within the target area. The dataset contains high-precision position coordinate sequences, velocity vector sequences, heading angle sequences, and millisecond-level timestamps of real targets. Simultaneously, acquire the original detection point data of the radar system in the same time period for the area. The point data includes the amplitude, signal-to-noise ratio (SNR), range, azimuth, and pitch measurements of the original video echo. Align the real target track data with the original radar detection point data in the time dimension, calculate the three-dimensional Euclidean distance between each real target position and the radar detection point in the spatial dimension based on the nearest neighbor algorithm, and establish track-point association pairs. Set a distance threshold and a velocity threshold. When the Euclidean distance is less than the set threshold and the velocity vector angle is less than the set angle, it is determined that the real target position has been successfully detected by the radar and marked as a "detection success event". Otherwise, it is marked as a "missed event". Based on the location coordinates of all real targets, combined with digital elevation model (DEM) data and land cover type data, the visibility between the radar and the target, the obstruction angle, and the multipath reflection coefficient are calculated. Based on successful detection events, missed detection events, line-of-sight results, radar system nominal parameters, and ambient noise floor, a binomial distribution statistical model is used to calculate the radar detection probability and measurement accuracy values ​​at different azimuth angles, elevation angles, and range thresholds. The calculated detection probability and accuracy values ​​are mapped to a geographic coordinate system and corrected using an atmospheric attenuation model to generate a radar detection coverage map and an accuracy distribution map.

2. The method for analyzing radar detection power based on target trajectory big data according to claim 1, characterized in that, After obtaining the large dataset of historical target tracks within the target area, the following data cleaning and coordinate system transformation steps are performed: The real target trajectory data is smoothed using a Kalman filter based on a uniform motion model. The process noise covariance matrix Q is set to a diagonal matrix, and the following covariance prediction formula is executed: ； in, The prior estimate is the error covariance matrix; This is the state transition matrix; The posterior estimation error covariance matrix; This is the process noise covariance matrix, whose diagonal elements are the variances of the position components. , , and velocity component variance , , ; For the raw radar detection point data, the instantaneous signal-to-noise ratio (SNR) is first calculated, and then points with an SNR less than 13 dB are removed. The SNR calculation formula is as follows: ； in, This refers to the peak power of the video pulse. This refers to the receiver's thermal noise power. The cleaned real target WGS-84 latitude and longitude coordinates (B, L, H) are transformed with the radar station center coordinate system (0, 0, 0) using the following coordinate transformation formula: ； in,( () represents the coordinates in the northeast-northeast coordinate system; This is a 3x3 rotation matrix calculated from the latitude and longitude of the radar station; The semi-major axis of the Earth is taken as 6,378,137 meters. Earth's eccentricity; ( (B, L, H) represents the coordinates of the radar station in the geocentric coordinate system; (B, L, H) represents the latitude, longitude, and elevation of the target point.

3. The method for analyzing radar detection power based on target trajectory big data according to claim 2, characterized in that, The establishment of track-point association pairs specifically includes: With radar scanning cycle Based on this, a time tolerance window is constructed. ; If a real target track point exists within the window, the radar scan time is calculated using cubic spline interpolation. The corresponding precise location of the target ( ); Calculate the spherical coordinate distance difference between the interpolated position and the radar mark. , , ; Construct the association cost function C and perform the following calculation: ； in, This represents the distance difference. This is the difference in azimuth; This is the pitch difference; For radar range resolution; This refers to the radar azimuth resolution. For radar elevation resolution; The weighting coefficients are satisfied. ; When the cost function C is less than the preset threshold of 1.5, the association is confirmed to be successful; otherwise, it is considered as unassociated.

4. The method for analyzing radar detection power based on target trajectory big data according to claim 3, characterized in that, The setting of the distance threshold and determination of the detection status also includes logic for eliminating "undetectable areas": Construct a radar resolution cell grid with a grid size of 500 meters in the range direction and 1 degree in the azimuth direction; For each real target location, calculate its corresponding radar beam center pointing angle; If the visibility calculation result of the actual target location is that it is blocked, that is, the line connecting the radar and the target intersects with the terrain elevation surface, it is directly marked as an "undetectable event" and is not included in the denominator of the detection probability statistics. If the actual target location is in a line-of-sight state but no matching point is found, and there are no clutter points within 5 resolution cells around the location, it is marked as a "missed event". If the actual target location is in a line-of-sight state but no matching point is found, and there are high-intensity clutter points around it (i.e., the clutter power exceeds the noise floor by 20dB), it is marked as a "masking and missed event" and is counted separately but not included in the radar's inherent detection performance. The formula for calculating the difference between clutter power and noise floor is as follows: ； in, clutter power; The noise floor power; the judgment condition is: >20dB.

5. The method for analyzing radar detection power based on target trajectory big data according to claim 4, characterized in that, The calculation of the line-of-sight and obstruction angle between the radar and the target specifically includes: On the line connecting the radar station and the target point, N terrain profile points are sampled at equal intervals of no more than 100 meters. Ground elevation of each sampling point was extracted from DEM data using bilinear interpolation. ; Calculate the radar line of sight at the sampling point The theoretical height Execute the following formula: ； in, This refers to the elevation of the radar antenna. Target elevation; The slant range of the sampling point from the radar; This represents the total slant range between the radar and the target. Calculate the radius of the first Fresnel zone of the radar beam Execute the following formula: ； in, The operating wavelength of the radar; If there are arbitrary sampling points satisfy > If so, the target is determined to be in the radar line-of-sight blind zone, and the shielding angle is calculated. : ； in, It is the shading angle.

6. The method for analyzing radar detection power based on target trajectory big data according to claim 5, characterized in that, The calculation of radar detection probability values ​​at different azimuth, elevation, and range angles using a binomial distribution statistical model specifically includes: The target airspace is divided into M azimuth sectors and N range gate units to form an M×N two-dimensional statistical matrix, where the azimuth resolution is 1 degree and the range gate width is the distance corresponding to the radar pulse width. For each two-dimensional grid cell Count the total number of real targets that fall into this unit. ; Count the number of targets that fall into this cell and are marked as "detection successful events". ; Calculate the detection probability of this cell. : ； for For small sample units, the probability values ​​are corrected using Bayesian estimation, and the following formula is applied: ； in, =1, =1 is a priori parameter; Simultaneously calculate the confidence interval for this probability value. , The confidence level is set at 95%, and the calculation formula is as follows: ； in, The quantiles of the standard normal distribution are given at a 95% confidence level. =1.

96.

7. The method for analyzing radar detection power based on target trajectory big data according to claim 6, characterized in that, The specific steps for analyzing radar measurement accuracy are as follows: Extract the actual target location coordinates from the "successful detection event". radar reported point coordinates ; Transform the coordinates to the radar station's northeast-sky coordinate system; Calculate range error : ； in, This refers to the range error; Calculate azimuth error : ； in, This refers to the azimuth error; Calculate the standard deviation of the range vector error within each range-azimuth grid. and the standard deviation of azimuth error : ； ； Where N is the number of samples in the grid; and These are the mean values ​​of distance and azimuth errors, respectively. And and This serves as the precision boundary value for the grid.

8. The method for analyzing radar detection power based on target trajectory big data according to claim 7, characterized in that, The generation of the radar detection coverage map includes the following visualization and data post-processing steps: Obtain the detection probability and accuracy values ​​of all calculated grid cells; The Gaussian kernel density estimation method is used to smooth the discrete grid data, and the kernel function bandwidth h is set to 2 grid cell widths. The probability contour levels were set to 0.1, 0.5, and 0.9, and the closing polygon boundaries corresponding to the probability values ​​were extracted using the Marching Squares algorithm. Closed boundaries are overlaid onto the electronic map layer, and map tile data is stored using a quadtree structure. Color values ​​are assigned based on probability values. The color mapping table uses the HSV color space, where the H component gradually changes from 0 degrees to 120 degrees, the S component is fixed at 100%, and the V component is negatively correlated with the precision value; the lower the precision, the lower the brightness.

9. The method for analyzing radar detection power based on target trajectory big data according to claim 8, characterized in that, After generating the radar detection coverage map, the process also includes model verification and iteration steps based on the KS test: Target trajectory data from the validation period that was not involved in modeling were selected as the test set, with a sample size of no less than 20% of the total data volume. Input the real target locations from the test set into the generated power coverage map, and read the theoretical detection probability for the corresponding locations. ; Set detection threshold Construct a confusion matrix if And if the radar actually detects a target, it is recorded as a true positive (TP); if However, the radar failed to detect the target, which was recorded as a false negative (FN). And if the radar does not detect the target, it is recorded as a true negative TN; if However, the radar detected the target and recorded it as a false positive (FP). Calculate the KS statistic ,in To detect the probability distribution function of successful samples, Let be the probability distribution function of the missed samples; If the KS statistic If the value is greater than 0.2, the model distribution difference is considered significant, and the distance threshold parameter needs to be adjusted and the steps repeated until... The value is less than or equal to 0.

2.

10. A radar detection capability analysis system based on target trajectory big data, used to implement the radar detection capability analysis method based on target trajectory big data as described in any one of claims 1-9, characterized in that, include: The acquisition module is used to acquire a large dataset of historical target tracks within the target area. The dataset includes a high-precision sequence of position coordinates, velocity vector, heading angle, and millisecond-level timestamps of real targets. At the same time, it acquires the original detection point data of the radar system in the area within the same time period. The point data includes the amplitude, signal-to-noise ratio (SNR), range, azimuth, and pitch measurements of the original video echo. The module is used to align real target track data with radar original detection point data in the time dimension, calculate the three-dimensional Euclidean distance between each real target position and radar detection point in the spatial dimension based on the nearest neighbor algorithm, and establish track-point association pairs. The determination module is used to set a distance threshold and a velocity threshold. When the Euclidean distance is less than the set threshold and the velocity vector angle is less than the set angle, the true target position is determined to be successfully detected by the radar and marked as a "detection success event". Otherwise, it is marked as a "missed event". The first calculation module is used to calculate the line-of-sight between the radar and the target, the obstruction angle, and the multipath reflection coefficient based on the location coordinates of all real targets, combined with digital elevation model (DEM) data and land cover type data. The second calculation module is used to calculate the detection probability value and measurement accuracy value of the radar under different azimuth angles, elevation angles and range gates based on successful detection events, missed detection events, line-of-sight results, radar system nominal parameters and environmental noise floor, using a binomial distribution statistical model. The generation module maps the calculated detection probability value and accuracy value to a geographic coordinate system, and corrects them by combining an atmospheric attenuation model to generate a radar detection power coverage map and an accuracy distribution map.

Images