Base station radar cooperative calibration method based on space-time synchronization and three-dimensional grid interpolation

Abstract

The application discloses a base station radar cooperative calibration method based on space-time synchronization and three-dimensional grid interpolation, and relates to the technical field of meteorological radar detection. The method comprises the following steps: acquiring original detection data of a 5G-A base station and a meteorological radar, performing coordinate unification and displacement compensation in the time dimension, and generating a time-synchronized virtual base station point cloud; constructing a k-d tree index, combining an antenna radiation pattern gain to calculate a composite weight, and mapping the virtual base station point cloud into a three-dimensional regular grid reflectivity matrix; combining base station parameters and local texture features to perform screening, and generating an effective sample mask; extracting effective interpolation reflectivity values according to the mask, calculating system deviation by distance segmentation and fitting a correction function, and generating updated base station radar constants. The application solves the problems of non-synchronization of the 5G-A base station and the meteorological radar in the space-time dimension and the measurement deviation caused by non-uniform beam filling, and realizes high-precision and adaptive calibration of the base station radar constants.

Description

Technical Field

[0001] This invention relates to the field of meteorological radar detection technology, and in particular to a base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional grid interpolation. Background Technology

[0002] With the development of 5G-Advanced (5G-A) technology, integrated sensing and communication base stations have acquired the ability to detect meteorological conditions (such as rainfall). However, compared to professional weather radars, 5G-A base stations are still immature in meteorological detection, and their detection accuracy is affected by hardware characteristics, signal attenuation, and algorithm model errors. In order for 5G-A base stations to output professional-grade meteorological data, they need to be calibrated using a standard weather radar as the "true value".

[0003] Existing calibration methods typically face the following technical challenges:

[0004] 1) Spatiotemporal asynchrony: The scanning times of 5G-A base stations and weather radars are different, and meteorological targets such as raindrops are in motion, which can lead to large errors when directly compared.

[0005] 2) Inconsistent spatial benchmarks: 5G-A base stations are based on local sector scanning, while weather radars are based on polar coordinate scanning. The data structures of the two are very different, making direct point-to-point comparison difficult.

[0006] 3) Beam non-uniformity: 5G-A antenna beams are directional, with huge differences in gain between the beam center and the edge, and are easily affected by the beam incomplete filling effect.

[0007] 4) Distance attenuation difference: The attenuation characteristics of 5G-A signals (especially millimeter waves) in rain are different from those of weather radar, and traditional single constant calibration cannot correct the system deviation that varies with distance.

[0008] Therefore, there is an urgent need for a high-precision collaborative calibration method that can eliminate spatiotemporal bias, correct beam effects, and fit distance-related biases. Summary of the Invention

[0009] This invention provides a base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional grid interpolation. This invention solves the problems of spatiotemporal asynchrony, inconsistent spatial references, beam non-uniformity, and range attenuation differences in existing technologies.

[0010] This invention provides a base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional grid interpolation, the method comprising:

[0011] The raw detection data from 5G-A base stations and weather radars are acquired, coordinates are unified, and displacement compensation in the time dimension is performed based on the echo motion vector to generate a time-synchronized virtual base station cloud.

[0012] Using a virtual base station cloud, a kd-tree index is constructed, and combined with the antenna radiation pattern gain to calculate the composite weight, the virtual base station cloud is mapped into a unified three-dimensional regular gridded reflectivity matrix.

[0013] Using a three-dimensional regular meshed reflectivity matrix as the processing object, and combining base station parameters and local texture features for filtering, an effective sample mask is generated that marks the effective contrast area.

[0014] Based on the effective sample mask, the effective interpolated reflectance values ​​are extracted from the three-dimensional regular gridded reflectance matrix. The system deviation is calculated by distance segment and a correction function is fitted to generate the updated base station radar constant.

[0015] The technical solution provided in this application has at least the following beneficial effects:

[0016] By calculating the echo motion vector using the optical flow method, accurate extrapolation of 5G-A base station data to weather radar time was achieved, eliminating dynamic errors caused by time differences. Three-dimensional mesh interpolation accelerated by kd-trees, combined with antenna pattern gain weighting, effectively solved the spatial matching problem of heterogeneous data sources and corrected the effects of beam non-uniformity. Through dual screening of texture features and engineering parameters, clutter, non-uniform echoes, and invalid geometric regions were eliminated, ensuring high reliability of the samples used for calibration. The single constant calibration mode was abandoned; by fitting a distance-bias function, both fixed system bias and distance-varying attenuation bias could be corrected simultaneously, significantly improving detection accuracy across the entire range.

[0017] In one optional implementation, raw detection data from 5G-A base stations and weather radar are acquired, coordinates are unified, and displacement compensation in the time dimension is performed based on echo motion vectors to generate a time-synchronized virtual base station cloud, including:

[0018] The raw detection data of 5G-A base station and the raw detection data of weather radar are read and preprocessed to obtain the preprocessed standard detection data of 5G-A base station and standard detection data of weather radar.

[0019] Using the standard Mocarto projection formula, the geographic coordinates of the standard meteorological radar detection data are projected onto a local tangent plane rectangular coordinate system so as to perform spatial operations with the standard detection data of 5G-A base stations and obtain the meteorological radar point cloud located in a unified local Cartesian coordinate system.

[0020] Using the spherical trigonometry formula and the Moccato projection formula, combined with the location of the 5G-A base station itself, the local spherical coordinates of the standard detection data of the 5G-A base station are transformed to the same Cartesian coordinate system as the meteorological radar point cloud to obtain the original point cloud of the 5G-A base station.

[0021] Based on the radar detection timestamp in the original detection data of the weather radar, the original point cloud of the 5G-A base station in the frame before and after the reference is selected.

[0022] The original point clouds of the 5G-A base station in the previous and next frames are rasterized into a front two-dimensional reflectivity intensity map and a back two-dimensional reflectivity intensity map, respectively.

[0023] Using the Lucas-Kanade optical flow method, the motion vector of each pixel in the front two-dimensional reflectivity intensity map and the back two-dimensional reflectivity intensity map is calculated to obtain the two-dimensional motion vector field covering the detection area.

[0024] Based on the two-dimensional motion vector field, the original point cloud of the 5G-A base station is extrapolated to the detection time of the weather radar, eliminating the time difference and obtaining a time-synchronized virtual base station point cloud.

[0025] In one optional implementation, the raw detection data of the 5G-A base station includes slant range, azimuth angle, elevation angle, reflectivity, and base station detection timestamp;

[0026] The raw detection data from the weather radar includes longitude, latitude, altitude, reflectivity factor, and radar detection timestamp.

[0027] In one optional implementation, the original point cloud of the 5G-A base station is extrapolated to the detection time of the weather radar based on the two-dimensional motion vector field, eliminating the time difference and obtaining a time-synchronized virtual base station cloud, including:

[0028] For each 5G-A base station in the previous frame, the corresponding motion vector is found in the two-dimensional motion vector field based on its planar coordinates;

[0029] Calculate the time shift ratio based on the radar detection timestamps of the previous and next frames.

[0030] Based on the time shift ratio, spatial displacement compensation is performed on each 5G-A base station in the previous frame to generate corresponding virtual points.

[0031] All virtual points are combined, and the reflectivity from the original detection data of the 5G-A base station is added to the corresponding virtual points to generate a time-synchronized virtual base station cloud.

[0032] In one alternative implementation, a virtual base station cloud is used to construct a kd-tree index. Combined with antenna radiation pattern gain, a composite weight is calculated, mapping the virtual base station cloud into a unified three-dimensional regular gridded reflectivity matrix, including:

[0033] Define a three-dimensional cubic space containing all the probe data, divide it into equally spaced voxels, and generate an initial three-dimensional regular mesh matrix framework including several meshes.

[0034] A kd-tree index is constructed using the coordinates of each virtual point in the virtual base station cloud as the key.

[0035] For each 5G-A base station detection point, calculate its deviation from the main beam direction of the base station antenna, obtain the corresponding gain value, and generate the pattern gain factor for each base station detection point.

[0036] Based on the kd-tree index, the effective voxel centers of each grid in the three-dimensional regular grid matrix framework are traversed to find the base station detection points around them, and the set of neighborhood points corresponding to each effective voxel center and the set of distances to the effective voxel center are generated.

[0037] By combining the distance set and the radiation pattern gain factor, the contribution weight of each neighboring point in the neighborhood point set to the effective voxel center is calculated, and the composite weight value of each base station detection point in the neighborhood is generated.

[0038] Based on the composite weight value, a weighted average is performed on the set of neighborhood points of each grid to generate the corresponding interpolated reflectance value;

[0039] The interpolated reflectance values ​​of all grids are filled into a three-dimensional regular grid matrix framework. Grids without neighboring points are left empty or have a specific background value, resulting in a three-dimensional regular gridded reflectance matrix.

[0040] In one optional implementation, for each base station detection point corresponding to a 5G-A base station, the degree of deviation relative to the main beam direction of the base station antenna is calculated, and the corresponding gain value is obtained to generate the radiation pattern gain factor corresponding to each base station detection point, including:

[0041] Calculate the position vector angle of each base station detection point relative to the 5G-A base station;

[0042] Calculate the normalized gain in this direction based on the base station antenna parameters of the corresponding 5G-A base station;

[0043] Based on the normalized gain and position vector angle, a simplified Gaussian pattern model is used to calculate the pattern gain factor for each base station detection point.

[0044] In one optional implementation, a three-dimensional regular meshed reflectivity matrix is ​​used as the processing object. This is combined with base station parameters and local texture features for filtering to generate an effective sample mask that marks the effective contrast areas, including:

[0045] Based on physical detection capabilities, invalid weak signals and unreliable far-field or near-field signals in the three-dimensional regular gridded reflectivity matrix are eliminated to generate a binary mask matrix of the preliminary effective area.

[0046] Calculate the local statistical features of the three-dimensional regular meshed reflectivity matrix to identify rainfall areas with uniform echoes, and generate a local texture feature matrix of the same size as the three-dimensional regular meshed reflectivity matrix;

[0047] By utilizing base station parameters and local texture feature matrices, regions with incomplete beam filling or severe geometric distortion in the three-dimensional regular meshed reflectivity matrix are removed to generate a binary mask matrix based on geometric effectiveness.

[0048] The binary mask matrix of the initial effective region, the binary mask matrix based on geometric validity, and the texture conditions are fused to generate an effective sample mask that marks the effective contrast region.

[0049] In one optional implementation, effective interpolated reflectance values ​​are extracted from the three-dimensional regular meshed reflectance matrix based on the effective sample mask. The system bias is calculated segmented by distance, and a correction function is fitted to generate updated base station radar constants, including:

[0050] By performing the same gridding mapping on the meteorological radar point cloud corresponding to the current three-dimensional regular gridded reflectivity matrix, a meteorological radar gridded reflectivity matrix with the same structure as the three-dimensional regular gridded reflectivity matrix is ​​obtained.

[0051] Based on the effective sample mask, the effective interpolated reflectance values ​​of the base station and the radar are extracted from the three-dimensional regular gridded reflectance matrix and the corresponding meteorological radar gridded reflectance matrix, respectively.

[0052] Calculate the spatial distance between the corresponding base station effective interpolated reflectance value and the radar effective interpolated reflectance value, and take the base station effective interpolated reflectance value, the radar effective interpolated reflectance value and the spatial distance as a valid sample. Collect several valid samples to form a valid sample dataset containing reflectance pairs and distance information.

[0053] The effective sample dataset is divided into multiple intervals according to spatial distance in order to analyze the systematic bias characteristics under different spatial distances and generate several interval sample sets sorted by spatial distance.

[0054] Perform statistical analysis on the interval sample set within each distance segment, calculate the average deviation, and generate the corresponding set of distance-deviation discrete data pairs.

[0055] Set a univariate linear regression model as the fitting model, use the least squares method with the average deviation as input, solve for the model parameters of the univariate linear regression model, minimize the sum of squared errors, and obtain the optimal model parameters;

[0056] Based on the radar meteorological equation and combined with a univariate linear regression model with optimal model parameters, the updated base station radar constants are obtained and written into the configuration file of the 5G-A base station. Attached Figure Description

[0057] Figure 1 The flowchart illustrates the steps of a base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional grid interpolation, as provided in an embodiment of the present invention. Detailed Implementation

[0058] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0059] The present invention will be further described below with reference to the accompanying drawings.

[0060] Reference Figure 1 The present invention provides a base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional grid interpolation, the method comprising:

[0061] S1: Acquire the raw detection data from 5G-A base stations and weather radar, unify the coordinates, and perform displacement compensation in the time dimension based on the echo motion vector to generate a time-synchronized virtual base station cloud.

[0062] S2: Using a virtual base station cloud, construct a kd-tree index, combine it with the antenna radiation pattern gain to calculate the composite weight, and map the virtual base station cloud into a unified three-dimensional regular gridded reflectivity matrix;

[0063] S3: Using a three-dimensional regular meshed reflectivity matrix as the processing object, and combining base station parameters and local texture features for filtering, an effective sample mask is generated to mark the effective contrast area.

[0064] S4: Extract effective interpolated reflectance values ​​from the three-dimensional regular gridded reflectance matrix based on the effective sample mask, calculate the system deviation by distance segment and fit the correction function to generate the updated base station radar constant.

[0065] The technical solution provided in this application has at least the following beneficial effects:

[0066] By calculating the echo motion vector using the optical flow method, accurate extrapolation of 5G-A base station data to weather radar time was achieved, eliminating dynamic errors caused by time differences. Three-dimensional mesh interpolation accelerated by kd-trees, combined with antenna pattern gain weighting, effectively solved the spatial matching problem of heterogeneous data sources and corrected the effects of beam non-uniformity. Through dual screening of texture features and engineering parameters, clutter, non-uniform echoes, and invalid geometric regions were eliminated, ensuring high reliability of the samples used for calibration. The single constant calibration mode was abandoned; by fitting a distance-bias function, both fixed system bias and distance-varying attenuation bias could be corrected simultaneously, significantly improving detection accuracy across the entire range.

[0067] In one optional implementation, raw detection data from 5G-A base stations and weather radar are acquired, coordinates are unified, and displacement compensation in the time dimension is performed based on echo motion vectors to generate a time-synchronized virtual base station cloud, including:

[0068] S11: Read the raw detection data of the 5G-A base station and the raw detection data of the weather radar respectively, and perform preprocessing to obtain the preprocessed standard detection data of the 5G-A base station and the standard detection data of the weather radar.

[0069] S12: Using the standard Mocarto projection formula, the geographic coordinates of the standard meteorological radar detection data are projected onto a local tangent plane rectangular coordinate system for spatial calculation with the 5G-A base station standard detection data, resulting in a meteorological radar point cloud located in a unified local Cartesian coordinate system. The formula is as follows:

[0070]

[0071] In the formula, This refers to the X-axis coordinates (usually representing the east-west direction, with east being positive) of the meteorological radar point cloud in the transformed Cartesian coordinate system. This refers to the Y-axis coordinates (usually representing the north-south direction, with north being positive) of the weather radar point cloud in the transformed Cartesian coordinate system. The coordinates of the point cloud of the weather radar in the transformed Cartesian coordinate system (representing vertical height and altitude); The longitude of the weather radar detection point; The latitude of the weather radar detection point; The longitude of the central meridian (i.e., the longitude corresponding to the local coordinate origin); Reference latitude (i.e., the latitude corresponding to the local coordinate origin); The average radius of the Earth is usually taken as the semi-major axis radius of the WGS-84 ellipsoid, which is approximately 6,378,137 meters (or 6,371,000 meters, depending on the specific accuracy requirements). The altitude of the weather radar detection point;

[0072] In this embodiment, the geographical location of base station BS_001 is selected as the origin; =116.391° (East Longitude); =39.907° (North Latitude); Longitude of Weather Radar Point Clouds ∈[116.35°, 116.43°], latitude of weather radar point cloud ∈[39.87°, 39.94°], Weather radar point cloud height ∈[0,6000] meters;

[0073] S13: Using the spherical trigonometry formula and the Moccato projection formula, combined with the location of the 5G-A base station, the local spherical coordinates of the standard detection data of the 5G-A base station are transformed to the same Cartesian coordinate system as the weather radar point cloud, to obtain the original point cloud of the 5G-A base station. The formula is as follows:

[0074]

[0075] In the formula, This represents the projected distance of the 5G-A base station detection point on the ground. The slant range detected by the 5G-A base station; The elevation angle of the 5G-A detection point; The absolute altitude of the target point (echo point); The altitude of the 5G-A base station antenna; the projection distance Combined with base station azimuth angle Convert to local tangent plane coordinates, then combine with reference latitude and longitude ( , Perform a Moccato projection to obtain coordinates in a unified Cartesian coordinate system. ); The coordinates of the original point cloud of the 5G-A base station in the transformed Cartesian coordinate system are the X-axis coordinates (usually representing the east-west direction, with east being positive). The Y-axis coordinate (usually representing the north-south direction, with north being positive) of the original point cloud of the 5G-A base station in the transformed Cartesian coordinate system; The coordinates of the Z-axis (representing vertical height and altitude) of the original point cloud of the 5G-A base station in the transformed Cartesian coordinate system;

[0076] S14: Based on the radar detection timestamp in the original detection data of the weather radar, select the original point cloud of the 5G-A base station in the frame before and the frame after the reference.

[0077] In this embodiment, the weather radar scanning period is 6 minutes (360 seconds), the base time is assumed to be 14:30:00, the 5G-A base station sampling rate is 1 second (i.e., outputting a point cloud snapshot once per second), and the selected time is:

[0078] The previous frame time is 14:29:00 (1 minute before the benchmark, to avoid nonlinear errors caused by excessive time span), and the next frame time is 14:31:00 (1 minute after the benchmark).

[0079] S15: Rasterize the original point cloud of the 5G-A base station in the previous and next frames into a front two-dimensional reflectivity intensity map and a back two-dimensional reflectivity intensity map, respectively.

[0080] S16: Using the Lucas-Kanade optical flow method, calculate the motion vector of each pixel in the front two-dimensional reflectivity intensity map and the back two-dimensional reflectivity intensity map to obtain the two-dimensional motion vector field covering the detection area.

[0081] S17: Based on the two-dimensional motion vector field, the original point cloud of the 5G-A base station is extrapolated to the detection time of the weather radar, eliminating the time difference and obtaining a time-synchronized virtual base station point cloud.

[0082] In one optional implementation, the raw detection data of the 5G-A base station includes slant range, azimuth angle, elevation angle, reflectivity, and base station detection timestamp;

[0083] The raw detection data from the weather radar includes longitude, latitude, altitude, reflectivity factor, and radar detection timestamp.

[0084] In one optional implementation, the original point cloud of the 5G-A base station is extrapolated to the detection time of the weather radar based on the two-dimensional motion vector field, eliminating the time difference and obtaining a time-synchronized virtual base station cloud, including:

[0085] S171: For each 5G-A base station in the previous frame, find the corresponding motion vector in the two-dimensional motion vector field based on its planar coordinates;

[0086] S172: Calculate the time shift ratio based on the radar detection timestamps of the previous and next frames, using the following formula:

[0087]

[0088] In the formula, As the base timestamp; The radar detection timestamp is the reference frame from the previous frame; The radar detection timestamp is the frame following the baseline. This represents the time shift ratio;

[0089] S173: Based on the time shift ratio, spatial displacement compensation is performed on each 5G-A base station in the previous frame to generate corresponding virtual points. The formula is as follows:

[0090]

[0091] In the formula, The virtual planar coordinates of the k-th echo point after time compensation; k is the echo point index. The virtual height coordinates of the k-th echo point after time compensation; The original coordinates of the k-th echo point in the original point cloud of the 5G-A base station in the previous frame; This represents the horizontal displacement component at the location of the echo point, derived from the motion vector. This represents the time shift ratio;

[0092] S174: Combine all virtual points and add the reflectivity from the original detection data of the 5G-A base station to the corresponding virtual points to generate a time-synchronized virtual base station cloud.

[0093] In one alternative implementation, a virtual base station cloud is used to construct a kd-tree index. Combined with antenna radiation pattern gain, a composite weight is calculated, mapping the virtual base station cloud into a unified three-dimensional regular gridded reflectivity matrix, including:

[0094] S21: Define a three-dimensional cubic space containing all the probe data, divide it into equally spaced voxels, and generate an initial three-dimensional regular mesh matrix framework including several meshes.

[0095] In this embodiment, the X-axis range is: −3000m to +3000m (6 km east-west).

[0096] Y-axis range: −3000m to +3000m (6 km in the north-south direction);

[0097] Z-axis range: 0m to 4000m (vertical height 4km).

[0098] S22: Construct a kd-tree index using the coordinates of each virtual point in the virtual base station cloud as the key value;

[0099] S23: For each 5G-A base station detection point, calculate its deviation from the main beam direction of the base station antenna, obtain the corresponding gain value, and generate the pattern gain factor for each base station detection point.

[0100] S24: Based on the kd-tree index, traverse the effective voxel centers of each grid in the three-dimensional regular grid matrix framework, find the base station detection points around them, and generate the set of neighborhood points corresponding to each effective voxel center and the set of distances to the effective voxel center.

[0101] S25: Combining the distance set and the radiation pattern gain factor, calculate the contribution weight of each neighboring point in the neighborhood point set to the effective voxel center, and generate the composite weight value of each base station detection point in the neighborhood, using the following formula:

[0102]

[0103] In the formula, is the composite weight value of the m-th base station detection point in the neighborhood; m is the index of the neighborhood point; is the Euclidean distance from the m-th neighboring point in the distance set to the center of the effective voxel; This is the distance attenuation exponent, typically taken as 2; is the pattern gain factor of the m-th base station detection point in the neighborhood; To set a minimum value (e.g., 1e−6) to prevent the weights from becoming ineffective when the gain is 0;

[0104] S26: Based on the composite weight value, a weighted average is calculated on the set of neighborhood points of each grid to generate the corresponding interpolated reflectance value. The formula is as follows:

[0105]

[0106] In the formula, For the i-th row, j-th column, and j-th column of a three-dimensional regular mesh matrix framework Interpolated reflectance results of the layered mesh; i is the row index of the 3D regular mesh; j is the column index of the 3D regular mesh; is the height layer index of the 3D regular mesh; m is the neighborhood point index; M is the total number of neighborhood points; is the composite weight value of the m-th base station detection point in the neighborhood; The original reflectance value of the m-th base station detection point in the neighborhood is derived from the virtual base station cloud.

[0107] S27: Fill the interpolated reflectance values ​​of all grids into the three-dimensional regular grid matrix framework. For grids without neighboring points, leave them empty or with a specific background value to obtain the three-dimensional regular gridded reflectance matrix.

[0108] In one optional implementation, for each base station detection point corresponding to a 5G-A base station, the degree of deviation relative to the main beam direction of the base station antenna is calculated, and the corresponding gain value is obtained to generate the radiation pattern gain factor corresponding to each base station detection point, including:

[0109] S231: Calculate the position vector angle of each base station detection point relative to the 5G-A base station;

[0110] S232: Calculate the normalized gain in this direction based on the base station antenna parameters of the corresponding 5G-A base station;

[0111] S233: Based on the normalized gain and position vector angle, a simplified Gaussian radiation pattern model is used to calculate the radiation pattern gain factor corresponding to each base station detection point. The formula is as follows:

[0112]

[0113] In the formula, The normalized pattern gain factor is the antenna gain intensity in the direction of the k-th echo point, and its value ranges from 0 to 1. When =1, it indicates that the point is located at the exact center of the beam (strongest energy). When ≈0, it indicates that the point is located at the edge of the beam or a side lobe (extremely weak energy); k is the echo point index; It is an exponential function; The horizontal angle between the k-th echo point and the central axis of the antenna main beam is derived from the position vector angle. The angle between the k-th echo point and the central axis of the antenna main beam is the vertical angle derived from the position vector angle. The vertical half-power beamwidth; This is the horizontal half-power beamwidth.

[0114] In one optional implementation, a three-dimensional regular meshed reflectivity matrix is ​​used as the processing object. This is combined with base station parameters and local texture features for filtering to generate an effective sample mask that marks the effective contrast areas, including:

[0115] S31: Based on physical detection capabilities, invalid weak signals and unreliable far-field or near-field signals in the three-dimensional regular gridded reflectivity matrix are eliminated to generate a binary mask matrix of the preliminary effective area.

[0116] In this embodiment, a signal strength threshold range is defined. , This represents the minimum signal strength threshold, corresponding to the base station receiver sensitivity (excluding noise). This represents the minimum signal strength threshold, corresponding to the saturation level (excluding distorted data).

[0117] Based on the location information in the base station's engineering parameters, calculate the slant distance from each grid center to the base station;

[0118] Define the effective detection range. , To minimize the effective detection distance, used to eliminate near-field blind zones and non-Rayleigh scattering regions. To effectively detect the maximum range, a sufficient signal-to-noise ratio must be ensured;

[0119] Based on the signal strength threshold range, slant range, and effective detection range, a preliminary binary mask matrix of the effective region is generated; the determination condition is that if the i-th row, j-th column, and j-th column of the three-dimensional regular grid matrix frame are within the range of the i-th column, j-th column, and j-th column, respectively... Interpolated reflectance results of layered mesh If the signal strength threshold is within the range and the corresponding slant range is within the effective detection range, then the i-th row, j-th column, and j-th column of the binary mask matrix are... The element of a layer is 1 if it is 1, otherwise it is 0.

[0120] S32: Calculate the local statistical features of the three-dimensional regular meshed reflectivity matrix to identify rainfall areas with uniform echoes and generate a local texture feature matrix of the same size as the three-dimensional regular meshed reflectivity matrix;

[0121] S33: Using base station parameters and local texture feature matrix, regions with incomplete beam filling or severe geometric distortion in the three-dimensional regular meshed reflectivity matrix are removed to generate a binary mask matrix based on geometric effectiveness.

[0122] In this embodiment, the azimuth and elevation angles corresponding to the grid points are calculated by using the latitude and longitude of the base station, and it is checked whether they are within the main lobe scanning range of the base station antenna (excluding the edge of the scanning sector).

[0123] Combined with the normalized pattern gain factor, calculate the average beam gain within the current grid area. If the average gain is too low (e.g., below the -3dB width range), it means that most of the beam did not hit the echo, which is "partial beam filling", which will lead to a serious underestimation and should be removed.

[0124] S34: The binary mask matrix of the initial effective region, the binary mask matrix based on geometric validity, and the texture conditions are fused to generate an effective sample mask that marks the effective contrast region.

[0125] In one optional implementation, effective interpolated reflectance values ​​are extracted from the three-dimensional regular meshed reflectance matrix based on the effective sample mask. The system bias is calculated segmented by distance, and a correction function is fitted to generate updated base station radar constants, including:

[0126] S41: The meteorological radar point cloud corresponding to the current three-dimensional regular gridded reflectivity matrix is ​​mapped using the same gridding method to obtain a meteorological radar gridded reflectivity matrix with the same structure as the three-dimensional regular gridded reflectivity matrix.

[0127] S42: Extract the effective interpolated reflectance values ​​of the base station and the radar from the three-dimensional regular gridded reflectance matrix and the corresponding meteorological radar gridded reflectance matrix, respectively, based on the effective sample mask;

[0128] S43: Calculate the spatial distance between the corresponding base station effective interpolated reflectivity value and the radar effective interpolated reflectivity value, and take the base station effective interpolated reflectivity value, the radar effective interpolated reflectivity value and the spatial distance as a valid sample. Collect several valid samples to form a valid sample dataset containing reflectivity pairs and distance information.

[0129] S44: Divide the effective sample dataset into multiple intervals according to spatial distance to analyze the systematic bias characteristics under different spatial distances and generate several interval sample sets sorted by spatial distance;

[0130] S45: Perform statistical analysis on the interval sample set within each distance segment, calculate the average deviation, and generate the corresponding set of distance-deviation discrete data pairs;

[0131] S46: Set a univariate linear regression model as the fitting model, use the least squares method with the average deviation as input, solve for the model parameters of the univariate linear regression model, and obtain the optimal model parameters by minimizing the sum of squared errors. The formula is:

[0132]

[0133] In the formula, The average deviation is given by a spatial distance of R; where R is the spatial distance. These are the model parameters for a univariate linear regression model, and the optimal model parameters are selected when the sum of squared errors is minimized.

[0134] S47: Based on the radar meteorological equations and combined with a univariate linear regression model with optimal model parameters, the updated base station radar constants are obtained and written into the configuration file of the 5G-A base station. The formula is as follows:

[0135]

[0136] In the formula, R is the reflectivity factor of the target object; R is the spatial distance. These are the original base station radar constants; This refers to the echo power received by the 5G-A base station receiver. Uncalibrated reflectivity measurements were calculated using the original base station radar constants for 5G-A base stations. The final, calibrated, accurate reflectivity value calculated using the updated base station radar constants for 5G-A base stations; These are the optimal model parameters.

[0137] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable hardware devices (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0138] The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (apparatus), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0139] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0140] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0141] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. "And / or" indicates that either one or both can be chosen. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes the element.

[0142] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional grid interpolation, characterized in that, The method includes: The raw detection data from 5G-A base stations and weather radars are acquired, coordinates are unified, and displacement compensation in the time dimension is performed based on the echo motion vector to generate a time-synchronized virtual base station cloud. Using a virtual base station cloud, a kd-tree index is constructed, and combined with the antenna radiation pattern gain to calculate the composite weight, the virtual base station cloud is mapped into a unified three-dimensional regular gridded reflectivity matrix. Using a three-dimensional regular meshed reflectivity matrix as the processing object, and combining base station parameters and local texture features for filtering, an effective sample mask is generated that marks the effective contrast area. Based on the effective sample mask, the effective interpolated reflectance values ​​are extracted from the three-dimensional regular gridded reflectance matrix. The system deviation is calculated by distance segment and a correction function is fitted to generate the updated base station radar constant.

2. The base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional mesh interpolation according to claim 1, characterized in that, The system acquires raw detection data from 5G-A base stations and weather radar, unifies coordinates, and performs time-dimensional displacement compensation based on echo motion vectors to generate a time-synchronized virtual base station cloud, including: The raw detection data of 5G-A base station and the raw detection data of weather radar are read and preprocessed to obtain the preprocessed standard detection data of 5G-A base station and standard detection data of weather radar. Using the standard Mocarto projection formula, the geographic coordinates of the standard meteorological radar detection data are projected onto a local tangent plane rectangular coordinate system so as to perform spatial operations with the standard detection data of 5G-A base stations and obtain the meteorological radar point cloud located in a unified local Cartesian coordinate system. Using the spherical trigonometry formula and the Moccato projection formula, combined with the location of the 5G-A base station itself, the local spherical coordinates of the standard detection data of the 5G-A base station are transformed to the same Cartesian coordinate system as the meteorological radar point cloud to obtain the original point cloud of the 5G-A base station. Based on the radar detection timestamp in the original detection data of the weather radar, the original point cloud of the 5G-A base station in the frame before and after the reference is selected. The original point clouds of the 5G-A base station in the previous and next frames are rasterized into a front two-dimensional reflectivity intensity map and a back two-dimensional reflectivity intensity map, respectively. Using the Lucas-Kanade optical flow method, the motion vector of each pixel in the front two-dimensional reflectivity intensity map and the back two-dimensional reflectivity intensity map is calculated to obtain the two-dimensional motion vector field covering the detection area. Based on the two-dimensional motion vector field, the original point cloud of the 5G-A base station is extrapolated to the detection time of the weather radar, eliminating the time difference and obtaining a time-synchronized virtual base station point cloud.

3. The base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional mesh interpolation according to claim 2, characterized in that, The raw detection data of the 5G-A base station includes slant range, azimuth angle, elevation angle, reflectivity, and base station detection timestamp; The raw detection data from the weather radar includes longitude, latitude, altitude, reflectivity factor, and radar detection timestamp.

4. The base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional mesh interpolation according to claim 3, characterized in that, Based on the two-dimensional motion vector field, the original point cloud of the 5G-A base station is extrapolated to the detection time of the weather radar, eliminating the time difference and obtaining a time-synchronized virtual base station point cloud, including: For each 5G-A base station in the previous frame, the corresponding motion vector is found in the two-dimensional motion vector field based on its planar coordinates; Calculate the time shift ratio based on the radar detection timestamps of the previous and next frames. Based on the time shift ratio, spatial displacement compensation is performed on each 5G-A base station in the previous frame to generate corresponding virtual points. All virtual points are combined, and the reflectivity from the original detection data of the 5G-A base station is added to the corresponding virtual points to generate a time-synchronized virtual base station cloud.

5. The base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional mesh interpolation according to claim 4, characterized in that, Using a virtual base station cloud, a kd-tree index is constructed. Combined with antenna radiation pattern gain, composite weights are calculated, mapping the virtual base station cloud into a unified three-dimensional regular gridded reflectivity matrix, including: Define a three-dimensional cubic space containing all the probe data, divide it into equally spaced voxels, and generate an initial three-dimensional regular mesh matrix framework including several meshes. A kd-tree index is constructed using the coordinates of each virtual point in the virtual base station cloud as the key. For each 5G-A base station detection point, calculate its deviation from the main beam direction of the base station antenna, obtain the corresponding gain value, and generate the pattern gain factor for each base station detection point. Based on the kd-tree index, the effective voxel centers of each grid in the three-dimensional regular grid matrix framework are traversed to find the base station detection points around them, and the set of neighborhood points corresponding to each effective voxel center and the set of distances to the effective voxel center are generated. By combining the distance set and the radiation pattern gain factor, the contribution weight of each neighboring point in the neighborhood point set to the effective voxel center is calculated, and the composite weight value of each base station detection point in the neighborhood is generated. Based on the composite weight value, a weighted average is performed on the set of neighborhood points of each grid to generate the corresponding interpolated reflectance value; The interpolated reflectance values ​​of all grids are filled into a three-dimensional regular grid matrix framework. Grids without neighboring points are left empty or have a specific background value, resulting in a three-dimensional regular gridded reflectance matrix.

6. The base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional mesh interpolation according to claim 5, characterized in that, For each 5G-A base station detection point, calculate its deviation from the main beam direction of the base station antenna, obtain the corresponding gain value, and generate the pattern gain factor for each base station detection point, including: Calculate the position vector angle of each base station detection point relative to the 5G-A base station; Calculate the normalized gain in this direction based on the base station antenna parameters of the corresponding 5G-A base station; Based on the normalized gain and position vector angle, a simplified Gaussian pattern model is used to calculate the pattern gain factor for each base station detection point.

7. The base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional mesh interpolation according to claim 6, characterized in that, Using a three-dimensional regular meshed reflectivity matrix as the processing object, and combining base station parameters and local texture features for filtering, effective sample masks that mark effective contrast areas are generated, including: Based on physical detection capabilities, invalid weak signals and unreliable far-field or near-field signals in the three-dimensional regular gridded reflectivity matrix are eliminated to generate a binary mask matrix of the preliminary effective area. Calculate the local statistical features of the three-dimensional regular meshed reflectivity matrix to identify rainfall areas with uniform echoes, and generate a local texture feature matrix of the same size as the three-dimensional regular meshed reflectivity matrix; By utilizing base station parameters and local texture feature matrices, regions with incomplete beam filling or severe geometric distortion in the three-dimensional regular meshed reflectivity matrix are removed to generate a binary mask matrix based on geometric effectiveness. The binary mask matrix of the initial effective region, the binary mask matrix based on geometric validity, and the texture conditions are fused to generate an effective sample mask that marks the effective contrast region.

8. The base station radar collaborative calibration method based on spatiotemporal synchronization and three-dimensional mesh interpolation according to claim 7, characterized in that, Based on the effective sample mask, effective interpolated reflectance values ​​are extracted from the three-dimensional regular meshed reflectance matrix. System bias is calculated segmented by distance, and a correction function is fitted to generate updated base station radar constants, including: By performing the same gridding mapping on the meteorological radar point cloud corresponding to the current three-dimensional regular gridded reflectivity matrix, a meteorological radar gridded reflectivity matrix with the same structure as the three-dimensional regular gridded reflectivity matrix is ​​obtained. Based on the effective sample mask, the effective interpolated reflectance values ​​of the base station and the radar are extracted from the three-dimensional regular gridded reflectance matrix and the corresponding meteorological radar gridded reflectance matrix, respectively. Calculate the spatial distance between the corresponding base station effective interpolated reflectance value and the radar effective interpolated reflectance value, and take the base station effective interpolated reflectance value, the radar effective interpolated reflectance value and the spatial distance as a valid sample. Collect several valid samples to form a valid sample dataset containing reflectance pairs and distance information. The effective sample dataset is divided into multiple intervals according to spatial distance in order to analyze the systematic bias characteristics under different spatial distances and generate several interval sample sets sorted by spatial distance. Perform statistical analysis on the interval sample set within each distance segment, calculate the average deviation, and generate the corresponding set of distance-deviation discrete data pairs. Set a univariate linear regression model as the fitting model, use the least squares method with the average deviation as input, solve for the model parameters of the univariate linear regression model, minimize the sum of squared errors, and obtain the optimal model parameters; Based on the radar meteorological equation and combined with a univariate linear regression model with optimal model parameters, the updated base station radar constants are obtained and written into the configuration file of the 5G-A base station.

Images