A method for identifying bird flocks and analyzing target differences in meteorological radar

Abstract

The present application discloses a method for weather radar bird flock identification and target difference analysis, which relates to the field of radar signal processing technology. The method includes preprocessing the weather radar raw data to obtain preprocessed weather radar data; screening the preprocessed weather radar data based on a first dynamic threshold method and a fuzzy logic classification method to obtain an initial bird flock target unit; screening the initial bird flock target unit based on spatiotemporal correlation analysis operation constraints to obtain a bird flock target; generating a bird flock activity distribution map based on the initial bird flock target unit and the bird flock target. Based on the bird flock activity distribution map, multidimensional feature extraction is performed on the preprocessed weather radar data to construct a multidimensional feature matrix of the bird flock target; principal component analysis, Fisher linear discriminant method, and clustering algorithm are used to calculate the multidimensional feature matrix of the bird flock and the weather target to obtain a target separability verification result; the present application realizes the accurate identification of bird flock targets.

Description

Technical Field

[0001] The present application relates to the field of radar signal processing technology, and in particular to a method for bird flock recognition and target difference analysis using weather radar. Background Art

[0002] With the rapid growth of civil aviation and the continuous improvement of the ecological environment, the overlap between low-altitude aircraft flight areas and bird activity areas has increased significantly, leading to a rising risk of bird strikes during takeoff and landing. This phenomenon is particularly prominent during migratory season, posing a severe challenge to aviation safety operations. Traditional bird prevention and control systems rely primarily on manual observation and basic detection methods, but they have significant shortcomings in terms of target coverage, monitoring accuracy, and real-time performance, making them unable to meet the needs of modern aviation safety protection.

[0003] In recent years, radar technology, leveraging its all-weather observation capabilities and automated monitoring advantages, has been widely used in bird ecology research. This has effectively overcome the technical bottlenecks of traditional manual observation, which are constrained by meteorological conditions and flight altitude restrictions. This has led to the development of two major technical approaches: specialized bird-detecting radar and meteorological radar. While specialized bird-detecting radar can detect and track small flocks or individual birds within a relatively small area, it faces application bottlenecks such as limited coverage and high networking costs. Meteorological radar, with its large-scale monitoring and networking advantages, has become a key tool for monitoring migratory bird flocks. However, its insufficient spatial resolution limits target recognition accuracy, and the high degree of aliasing between bird flocks and non-biological targets such as precipitation in the raw echo data exacerbates the technical difficulty of feature separation. Especially when dealing with low, slow, and small flying bird targets, current approaches rely primarily on deep feature learning and multidimensional feature models. Deep feature learning extracts salient features from the data, which are then combined with the target's physical properties and multidimensional feature models for comprehensive analysis. However, existing technology systems face dual challenges: on the one hand, deep learning-based feature extraction methods have inherent flaws such as weak correlation with physical mechanisms and strong data dependence; on the other hand, multidimensional feature models lack systematic parameterized representations, and the discriminative quantification of different feature parameters is insufficient, resulting in key discriminant features being easily overwhelmed in high-dimensional space, severely restricting the optimization efficiency of the classifier's decision boundary. Therefore, effectively overcoming key issues such as high data dependence and insufficient model interpretability in weather radar bird flock target recognition and feature extraction is of great significance. Summary of the Invention

[0004] The purpose of this application is to provide a weather radar bird flock recognition and target difference analysis method, which can achieve accurate identification of bird flock targets.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for bird flock recognition and target difference analysis using a weather radar, comprising:

[0007] Acquiring raw weather radar data and performing a preprocessing operation on the raw weather radar data to obtain preprocessed weather radar data;

[0008] Based on a first dynamic threshold method and a fuzzy logic classification method, the preprocessed weather radar data is screened to obtain an initial bird flock target unit; the first dynamic threshold method is a first threshold of a preset multidimensional feature; the initial bird flock target unit is multiple;

[0009] Based on the spatiotemporal correlation analysis operation constraint condition, the initial bird flock target unit is secondary screened to obtain the bird flock targets in the initial bird flock target unit;

[0010] generating a bird flock activity distribution map based on the initial bird flock target unit and the bird flock targets in the initial bird flock target unit;

[0011] Based on the bird flock activity distribution map, multi-dimensional feature extraction is performed on the pre-processed weather radar data to obtain multi-dimensional feature information of the bird flock target;

[0012] Constructing a bird flock target multidimensional feature matrix based on the bird flock target multidimensional feature information;

[0013] Obtaining a multi-dimensional feature matrix of meteorological targets;

[0014] The multidimensional feature matrix of the bird flock target and the multidimensional feature matrix of the meteorological target are processed using principal component analysis, Fisher linear discriminant method and clustering algorithm to obtain target separability verification results; the target separability verification results are used to quantify the separability of the bird flock target and the meteorological target.

[0015] In a second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method for weather radar bird flock identification and target difference analysis.

[0016] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for weather radar bird flock identification and target difference analysis.

[0017] In a fourth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned weather radar bird flock recognition and target difference analysis method.

[0018] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0019] The present application provides a method for weather radar bird flock identification and target difference analysis. By adopting the first dynamic threshold method, fuzzy logic classification method and spatiotemporal correlation analysis operation constraints to screen the bird flock targets present in the weather radar raw data, it is possible to effectively identify targets related to bird flock activities from the weather radar data, exclude other interference factors or non-bird targets, and ensure the accuracy of bird flock target identification. Moreover, it can avoid the problems of high data dependence and insufficient model interpretability when using deep feature learning and multidimensional feature models for target identification. In addition, by using principal component analysis, Fisher linear discriminant method and clustering algorithm to calculate the target separability verification results, it is possible to more accurately verify the separability of bird flock targets and meteorological targets, and improve the accuracy of target differentiation. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 This is an application environment diagram of a method for weather radar bird flock recognition and target difference analysis in one embodiment of the present application;

[0022] Figure 2 A schematic flow chart of a method for bird flock identification and target difference analysis using weather radar provided in one embodiment of the present application;

[0023] Figure 3 for Figure 2 Detailed flow diagram of step 201;

[0024] Figure 4 A schematic diagram of the overall process of a method for weather radar bird flock identification and target difference analysis provided in one embodiment of the present application;

[0025] Figure 5 A schematic diagram of the functional modules of a target difference analysis module provided in another embodiment of the present application;

[0026] Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0028] The limitations of existing technical solutions are primarily reflected in: 1) Traditional manual screening methods are inefficient and unreliable; 2) Deep feature learning is disconnected from target scattering mechanisms, resulting in poor interpretability of multidimensional feature models; and 3) The feature parameter system lacks quantitative characterization standards, making it difficult to construct a classification decision model with clear physical meaning. These shortcomings directly impact the accuracy and engineering applicability of bird flock target recognition, necessitating the urgent need for technological innovation to develop new solutions that deeply integrate physical mechanisms with feature engineering.

[0029] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0030] The weather radar bird flock recognition and target difference analysis method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the raw data of the weather radar to the server 104. For the raw data of the weather radar, the server 104 preprocesses the raw data of the weather radar to obtain the preprocessed weather radar data; based on the first dynamic threshold method and the fuzzy logic classification method, the preprocessed weather radar data is screened to obtain the initial bird flock target unit; based on the spatiotemporal correlation analysis operation constraint condition, the initial bird flock target unit is screened to obtain the bird flock target; based on the initial bird flock target unit and the bird flock target, a bird flock activity distribution map is generated. Based on the bird flock activity distribution map, multidimensional feature extraction is performed on the preprocessed meteorological radar data to construct a multidimensional feature matrix of the bird flock target; the multidimensional feature matrix of the bird flock and meteorological target is calculated using principal component analysis, Fisher linear discriminant method and clustering algorithm to obtain the target separability verification result; the server 104 can feed back the obtained target separability verification result to the terminal 102.

[0031] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, and IoT devices. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or a cloud server.

[0032] In an exemplary embodiment, Figure 2 As shown, a method for weather radar bird flock identification and target difference analysis is provided. The method is executed by a computer device, specifically a computer device such as a terminal or a server, or a terminal and a server. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used as an example to illustrate the process, including the following steps 201 to 208.

[0033] Step 201: obtain raw weather radar data and perform preprocessing operations on the raw weather radar data to obtain preprocessed weather radar data.

[0034] Step 202, based on the first dynamic threshold method and the fuzzy logic classification method, the preprocessed weather radar data is screened to obtain the initial bird flock target unit; the first dynamic threshold method is a first threshold of a preset multi-dimensional feature; the initial bird flock target unit is multiple.

[0035] Step 203 : Based on the spatiotemporal correlation analysis operation constraint condition, the initial bird flock target unit is screened a second time to obtain the bird flock targets in the initial bird flock target unit.

[0036] Step 204: Generate a bird flock activity distribution map based on the initial bird flock target unit and the bird flock targets in the initial bird flock target unit.

[0037] Step 205 : Based on the bird flock activity distribution map, multi-dimensional feature extraction is performed on the pre-processed weather radar data to obtain multi-dimensional feature information of the bird flock target.

[0038] Step 206: construct a bird flock target multidimensional feature matrix based on the bird flock target multidimensional feature information.

[0039] Step 207: Obtain a multi-dimensional feature matrix of meteorological targets.

[0040] In step 208, the multidimensional feature matrix of the bird flock target and the multidimensional feature matrix of the meteorological target are processed using principal component analysis, Fisher linear discriminant method, and clustering algorithm to obtain target separability verification results; the target separability verification results are used to quantify the separability of the bird flock target and the meteorological target.

[0041] By implementing the above-mentioned steps 201 to 208, by adopting the first dynamic threshold method, the fuzzy logic classification method and the spatiotemporal correlation analysis operation constraint conditions to screen the bird flock targets present in the raw data of the weather radar, it is possible to effectively identify targets related to bird flock activities from the weather radar data, exclude other interference factors or non-bird targets, and ensure the accuracy of bird flock target identification. Moreover, it is possible to avoid the problems of high data dependence and insufficient model interpretability when using deep feature learning and multidimensional feature models for target identification. In addition, by using principal component analysis, Fisher linear discriminant method and clustering algorithm to calculate the target separability verification results, it is possible to more accurately verify the separability of bird flock targets and meteorological targets, and improve the accuracy of target differentiation.

[0042] Furthermore, if Figure 3 As shown, in step 201, raw weather radar data is obtained and preprocessed to obtain preprocessed weather radar data. The specific process is as follows:

[0043] The weather radar used in this application operates in the S band and can theoretically provide meteorological observation information within a range of up to 450 kilometers, with multiple scanning modes. In dual-polarization operating mode, the weather radar can provide relatively rich observation information. Taking each volume scanning unit as the basic unit, the multi-dimensional feature information provided in each unit includes time, distance, azimuth, pitch angle, reflectivity, radial velocity, spectrum width, differential reflectivity, correlation coefficient, differential phase, etc. These features are key characteristic indicators in bird flock target detection and provide important data support for bird target identification, feature extraction, and analysis.

[0044] When preprocessing the raw data files of weather radars, since the radar near-field area is extremely susceptible to interference from ground clutter (such as mountains and buildings), the feature data is first cleaned and invalid values are eliminated by setting a threshold for multi-dimensional features (i.e., the second dynamic threshold method). At the same time, considering that different feature information may have dimensional differences, and such differences will significantly interfere with the accuracy and scientific nature of subsequent data analysis results. Based on this, this application further normalizes the cleaned data to obtain the elimination results; with the help of the normalization algorithm, various types of feature information are uniformly mapped to specific intervals, effectively eliminating the potential interference of dimensional differences on the feature space, achieving the same scale between features, and laying a solid foundation for subsequent data analysis and model construction.

[0045] Then, the data file reading package decompresses and parses the raw weather radar data files provided by the Meteorological Bureau. The data, including the latitude and longitude, azimuth and elevation angles, time, and secondary product data (i.e., multidimensional features, including reflectivity, radial velocity, spectral width, differential reflectivity, differential phase, and correlation coefficient) contained in each volume scan unit of the weather radar, is saved in a specified .mat format as a data file readable by the MATLAB environment. Each raw weather radar data file corresponds to a parsed MATLAB data file. Based on this, the .mat format files are used as input and analyzed for the temporal and spatial characteristics of the volume scan units and various product characteristics, thereby extracting the multidimensional characteristic information of the bird flock targets contained therein. The extracted multidimensional characteristic information of the bird flock targets is also stored in .mat format for subsequent research.

[0046] When reading weather radar raw data files, a structured binary decoding engine is first built in the data parsing layer to perform byte-level parsing of the radar file header and extract basic parameters. Subsequently, based on the dynamic memory allocation mechanism of the radial data stream, each radial block is read cyclically and physical quantity conversion is performed:

[0047] (1);

[0048] in, is the physical quantity after conversion; The original reflectivity / velocity data; offset is the offset, scale is a scaling factor, both of which are dynamically defined by the scanning strategy and are used for normalization of reflectivity (Z), velocity (V) and other data.

[0049] In the coordinate conversion stage, in order to solve the positioning error caused by the curvature of the earth, the 4 / 3 equivalent earth radius model is introduced, and its height correction formula is:

[0050] (2);

[0051] in, Indicates the corrected target altitude; Indicates the measured slope distance; represents the equivalent Earth radius; Indicates the installation height of the device.

[0052] At the same time, the mapping from polar coordinates to plane rectangular coordinates is achieved through orthogonal decomposition as follows:

[0053] (3);

[0054] in, Indicates the horizontal abscissa; Indicates the horizontal vertical coordinate; Indicates the vertical height of the target point; represents radial distance in polar coordinates; Represents the geometric correction parameters.

[0055] The longitude and latitude offsets are further calculated using the modified Newton (newll_dxdy) algorithm:

[0056] (4);

[0057] in, Indicates the longitude offset, Indicates the latitude offset, 、 Represents the plane rectangular coordinate offset, Represents the projection parameters, which are calibrated by the radar site coordinates and the projection parameters to generate three-dimensional geographic coordinates including longitude and latitude.

[0058] Meteorological radars primarily observe targets with large spatial scales, resulting in larger volumetric units. This volumetric unit size increases quadratically with increasing distance. This significantly differs from traditional high-resolution bird-detecting radars. Therefore, for meteorological radars, determining whether a flying bird target (also known as a flock of birds) is essentially a matter of determining whether a particular volumetric unit contains a bird target. The actual target characteristics of a volumetric unit, such as reflectivity, differential reflectivity, and correlation coefficient, are actually the coherent superposition of all scattered radar echoes within the corresponding volumetric unit. The target characteristics reflected by the volumetric unit make it difficult to determine whether a single bird or a flock of birds is present, and determining the species and distribution of birds is extremely challenging.

[0059] The current method used in this application to determine whether a certain scanning unit of a weather radar contains a bird target is divided into two stages. The first stage is to independently determine whether a certain scanning unit contains a bird target based on the target characteristics of the weather radar scanning unit. The second stage is to analyze whether the current scanning unit contains a bird target within a larger temporal and spatial scale based on the temporal and spatial distribution characteristics of birds, especially migratory birds.

[0060] Phase 1: Bird target identification method based on a single scanning unit.

[0061] Currently, bird detection methods based on single-body scanning units employ a supervised learning framework, extracting features from a large number of reliable samples and constructing a classification model through statistical modeling to achieve positive identification of unknown units. However, due to the lack of sufficient labeled samples in current weather radar data, directly applying such methods presents modeling difficulties.

[0062] To this end, further, in step 202 of the present application, based on the first dynamic threshold method and the fuzzy logic classification method, the pre-processed meteorological radar data is screened to obtain the initial bird flock target unit; specifically: based on the first dynamic threshold method, the non-bird flock target units in the pre-processed meteorological radar data are excluded to obtain the potential bird flock target unit; based on the fuzzy logic classification method, the target probability value of the potential bird flock target unit is calculated to obtain the target probability value corresponding to the potential bird flock target unit; based on the target probability value corresponding to the potential bird flock target unit, the potential bird flock target unit is screened to obtain the initial bird flock target unit. In the process of calculating the target probability value of the potential bird flock target unit based on the fuzzy logic classification method, the fuzzy logic classification method of the NEXRAD system is used for reference, combined with the characteristic response difference of the X / S band radar to the bird flock target, the initial bird flock target unit is screened, and the time-space correlation analysis is introduced to optimize the judgment result.

[0063] According to existing research results on the application of NEXRAD systems for biological target identification (birds, insects, bats), there are several key differences in the dual-polarization characteristics of radar targets between bird flocks and meteorological targets (rain, snow, hail, typhoons, etc.):

[0064] ① Reflectivity reflects the spatial distribution density of targets by quantifying the coherent summation intensity of backscattered echoes from scatterers within the radar scanning unit. Due to the differences in scattering characteristics between meteorological and biological targets, reflectivity effectively characterizes the degree of clustering of scatterers within a scanning unit. For example, the reflectivity of dense raindrop areas is significantly higher than that of sparse flying birds, making it one of the fundamental physical quantities for distinguishing target types.

[0065] (5);

[0066] in, represents the reflectivity factor; represents the radar cross section of a single scatterer; Indicates the wavelength of electromagnetic waves emitted by the radar; represents the dielectric factor; Represents the sixth power of the diameter of a single scatterer.

[0067] ② Radial velocity characterizes the statistical mean of the radial velocity of scatterers within the scanning unit based on the Doppler effect. The radial velocity of biological targets (such as birds) is affected by environmental factors and has a wide range of values, significantly overlapping with meteorological targets (such as precipitation). This limits its discriminatory power when used alone for target classification. Therefore, it is often used as an auxiliary parameter to analyze target motion trends and migration patterns.

[0068] (6);

[0069] in, represents radial velocity; represents the unit direction vector; represents the target velocity vector.

[0070] ③ Differential reflectivity is a key polarization parameter for distinguishing meteorological from biological targets. Its physical mechanism stems from the difference in scattering ability of a scatterer for horizontally and vertically polarized electromagnetic waves. For meteorological targets (such as raindrops and hail), the geometric shape of the scatterers is uniform and symmetrical, with similar horizontal and vertical polarization reflectivities and ZDR values approaching 0dB. However, for biological targets (such as birds), the horizontal polarization scattering is significantly stronger than the vertical polarization due to the inhomogeneity of the medium and their complex shape, resulting in a significantly positive ZDR value. This characteristic is closely related to the scatterer size (when it is close to or smaller than the radar wavelength) and geometric anisotropy, and can serve as an effective criterion for radar echo classification, especially for distinguishing between meteorological and biological targets based on the differences in polarization characteristics.

[0071] (7);

[0072] in, represents the differential reflectivity; represents the horizontal polarization reflectivity factor; represents the vertical polarization reflectivity factor.

[0073] ④ Differential phase characterizes the differences in target scattering properties through the phase difference between horizontally and vertically polarized reflected waves. Meteorological targets (such as raindrops and hailstones) have small inter-polarization phase differences due to their geometric symmetry. However, biological targets (such as birds) experience significantly larger and more complex phase differences due to their inhomogeneous media and complex shapes. Its range is influenced by multiple factors, including target geometry and velocity, and lacks clear boundaries. This parameter, in conjunction with differential reflectivity, can effectively distinguish between the polarization-based scattering characteristics of meteorological and biological targets.

[0074] (8);

[0075] in, represents the differential phase; represents the complex conjugate of the horizontally polarized scattered field; represents the vertically polarized scattered field; represents the horizontally polarized scattered field.

[0076] ⑤ The correlation coefficient effectively distinguishes meteorological and biological targets by quantifying the similarity of horizontally and vertically polarized radar echoes. Meteorological scatterers, due to their polarization insensitivity, have high echo similarity, resulting in a correlation coefficient approaching 1. However, for biological targets (such as birds), echo similarity is significantly reduced due to their polarization sensitivity and complex appearance. Statistics show that their correlation coefficients are concentrated in the range of 0.3-0.8. This parameter, in conjunction with differential reflectivity and differential phase, forms the core feature system for polarization radar identification of biological targets.

[0077] (9);

[0078] in, represents the correlation coefficient; represents the complex conjugate of the horizontally polarized scattered field; represents the vertically polarized scattered field; represents the horizontally polarized scattered field.

[0079] 6. Spectral width characterizes the standard deviation of the radial velocity of scatterers within the radar's volume scanning unit and reflects the dispersion of the velocity distribution. However, the radial velocity distributions of meteorological and biological scatterers overlap significantly and are susceptible to clutter interference, resulting in insufficient discrimination of spectral width parameters for distinguishing between biological and non-biological targets. In practical applications, a comprehensive discrimination is often required, combining polarization parameters such as reflectivity and differential reflectivity.

[0080] (10);

[0081] in, represents the spectrum width; represents the square of radial velocity; represents the radial velocity.

[0082] Based on the above analysis of the target characteristics of flying birds in the volume scanning unit of weather radar, combined with existing radar observation data and the target classification method used by the NEXRAD system, the method used in this application to determine whether a volume scanning unit contains a flying bird target is as follows:

[0083] For a certain meteorological radar volume scanning unit, based on the above analysis of the target characteristics of the volume scanning unit containing flying bird targets, a preliminary hard classification can be performed based on the value of a certain category feature contained in the volume scanning unit, that is, the volume scanning unit whose target feature value is not within the expected range is classified as a non-bird unit, and the target feature is any multidimensional feature. On this basis, the fuzzy logic classification method is used to construct a probability distribution model of different types of target features, and the target category judgment is made in combination with the weight allocation algorithm. Define the membership function For the Class target characteristics Corresponding to The probability of the target class. Category target, the Class target characteristics The corresponding probability value can be described by a trapezoidal function. For a certain unknown type of target’s polarization feature V={Vj|j=1,2,…,6}, its corresponding The comprehensive judgment probability (target probability value) of the class target is:

[0084] (11);

[0085] in, is the weight matrix element, indicating the Class target characteristics In the Importance of class target recognition. Based on the comprehensive judgment probability defined in formula (11), the category to which the body scanning unit belongs is determined to be the category with the highest corresponding probability density, thereby achieving a preliminary judgment of the body scanning unit category.

[0086] The second stage: data mining method based on the spatial and temporal correlation characteristics of volume scanning units.

[0087] Based on the above analysis, although the fuzzy logic classification method can filter out the volume scanning units of obvious non-bird targets based on the S-band radar observation data of the US NEXRAD system, its discrimination accuracy is limited by the single-dimensional feature analysis, and it is difficult to achieve accurate identification of units containing bird targets. Therefore, this application proposes a hierarchical progressive detection framework: first, a screening is performed through the characteristics of a single volume scanning unit, and then the spatiotemporal correlation analysis operation constraints are introduced to enhance the reliability of the discrimination. Specifically, the traditional method only relies on isolated parameters such as reflectivity thresholds, ignoring the correlation characteristics of the target in time evolution, spatial distribution and vertical structure, which is the key to distinguishing "low, slow and small" bird targets from meteorological clutter.

[0088] In meteorological radar observations, meteorological targets dominate, while bird targets, as non-primary targets that are "low, slow, and small", require weak signals to be extracted from a complex background for detection. In order to address the challenge that bird targets are easily misjudged as invalid units, a secondary screening is performed. Furthermore, in step 203, based on the spatiotemporal correlation analysis operation constraints, the initial bird flock target unit is secondary screened to obtain the bird flock targets in the initial bird flock target unit, including: processing the initial bird flock target unit based on a preset neighborhood density to obtain a neighborhood density screening result; processing the neighborhood density screening result based on a preset height dimension to obtain a height dimension screening result; processing the height dimension screening result based on a preset spatiotemporal continuity constraint to obtain a spatiotemporal correlation analysis result. Specifically: First, the bird flock appears "sparse" in the horizontal dimension, that is, there are few similar targets in the vicinity of the body scanning unit, and dense meteorological targets can be distinguished through density threshold and neighborhood analysis; second, its height distribution presents a "thin" feature, and the characteristics between elevation angles in the vertical direction suddenly change, which contrasts with the continuous and slow change of meteorological targets; finally, migratory bird flocks have a "fast" feature, and their flight speed far exceeds the static ground clutter. Combined with the spatiotemporal changes of adjacent 6-minute radar data, meteorological interference can be effectively filtered out; in addition, the flying bird body scanning unit presents a "small" characteristic, that is, the spatial distribution range is limited and the density is low, and the reflectivity, correlation coefficient and other characteristics are significantly different from those of the neighboring units. Based on the above characteristics, this application designs a four-level screening algorithm, which gradually eliminates non-bird targets through dynamic thresholds, neighborhood density, height dimension verification and spatiotemporal correlation, and finally extracts high-probability flying bird units for big data verification, significantly improving detection efficiency and aviation safety early warning capabilities.

[0089] Weather radar-based bird flock target identification relies on the rich feature information in the echo. Bird flocks exhibit unique characteristics in radar echoes, which can be distinguished from weather echoes, enabling flock detection and identification. Weather radar differs significantly from traditional bird-detecting radar in signal transmission format and scanning mode. Because the primary observation targets are large-scale meteorological targets, the volume scanning unit of a weather radar is large, and its volume increases quadratically with increasing distance. Traditional bird-detecting radars, on the other hand, focus on detecting small-scale biological targets and have higher resolution. Yantai, located in Shandong Province, China, lies on the East Asian-Australasian Flyway. As an important migratory route for migratory birds, large numbers of migratory birds pass through the area each spring and autumn. This application focuses on the collection and analysis of target data of migratory bird flocks on a large scale. The compiled dataset is mainly based on the Yantai migratory bird autumn migration period from September to December. The precipitation data on September 30, 2024, the clear sky data on October 3, 2024, and the clear sky and cloudy data on December 5, 2024 are mainly selected for research and construction of the dataset.

[0090] Furthermore, in step 208, principal component analysis (PCA), Fisher linear discriminant (FLD), and clustering algorithm are used to process the multidimensional feature matrix of the flock target and the multidimensional feature matrix of the meteorological target to obtain target separability verification results, which specifically include:

[0091] The multidimensional feature matrix of bird flock targets and meteorological targets was processed by principal component analysis to obtain the first dimensionality reduction feature space. The hydrometeor classification algorithm (HCA) was used to identify meteorological targets, and a multidimensional feature matrix was constructed for the identified meteorological targets.

[0092] Fisher linear discriminant method is used to process the multidimensional feature matrix of bird flock targets and the multidimensional feature matrix of meteorological targets to obtain the second dimensionality reduction feature space.

[0093] The clustering algorithm is used to process the first dimensionality reduction feature space and the second dimensionality reduction feature space to obtain the target separability verification result.

[0094] This application uses principal component analysis and Fisher linear discriminant method for feature analysis. Among them, the main purpose of principal component analysis is to reduce the data dimension and remove redundant information in the data. First, define the target feature matrix with the target feature vector as the column vector as ,in Represents a set of eigenvectors. The basic principle of subspace analysis is to decompose such an original feature matrix so that the elements in X can be described by a set of decorrelated basis vectors. This requires finding a mapping matrix W to project X into a new feature space S, that is:

[0095] (12);

[0096] Among them, S is the low-dimensional representation of the original data in the subspace, is the mapping matrix. Principal component analysis and Fisher linear discriminant method calculate the mapping matrix W based on different principles.

[0097] The purpose of principal component analysis is to find an optimal set of unit orthogonal vector bases through linear transformation, use their linear combination to reconstruct the original samples, and minimize the error between the reconstructed samples and the original samples. The basic principle of PCA in finding the mapping matrix W is to maximize the total scatter matrix of the entire feature sample set. The total scatter matrix is defined as follows:

[0098] (13);

[0099] in, is the average of all eigenvectors, For the k The feature vector of the samples, represents the transpose operation of a vector, N is the total number of samples. Mathematically, the calculation of the mapping matrix W is achieved by solving the following optimization problem:

[0100] (14);

[0101] in, represents the optimal solution, is the total scatterer matrix, is the mapping matrix. Physically speaking, principal component analysis projects high-dimensional data into a low-dimensional space through linear transformation, effectively removing redundant information from the data while preserving the data variance to the greatest extent possible. This allows the reduced-dimensional data to present the main structure and features of the original data in the new feature subspace in the most concise and critical way, thereby accurately reflecting the global distribution of the original feature space.

[0102] Unlike principal component analysis, Fisher linear differentiation method pays more attention to the relationship between different class elements when establishing the mapping matrix. In the mapping process, FLD optimizes the classification structure of different class data and enhances the separability of the classification interface in the feature space. In order to achieve this goal, we first need to establish the intra-class scatterers. and inter-class scatterers Concept:

[0103] (15);

[0104] (16);

[0105] in, is the total number of categories, It is The number of samples in the class, It is The first in the category sample vectors, is the global mean vector of all samples, It is The sample mean vector of the class.

[0106] From a physical point of view, intra-class scatterers describe the distance between all feature vectors of the same type and the center of the sample of that category. The smaller the value, the more compact the distribution of feature vectors of samples of the same type in the feature space; and the inter-class scatterers It describes the distance between different types of feature vectors in the feature space. The larger the , the more obvious the difference between the feature vectors of different categories in the feature space, which is more conducive to the establishment of the classification surface. Therefore, in order to achieve the best classification performance, it is necessary to ensure that the projection direction It can simultaneously maximize the inter-class scatterers and minimize the intra-class scatterers. At this time, the objective function is the largest, as shown in formula (16).

[0107] (17);

[0108] in, is the projection direction vector, represents the measure of inter-class differences after projection, A measure representing the intra-class variance after projection.

[0109] Based on the feature space after dimensionality reduction, the separability of the target feature is verified by quantifying the cluster center, coverage radius, and class spacing. The cluster center is defined as the mean vector of the feature vectors of the same sample:

[0110] (18);

[0111] in, For the No. The feature vector of the samples, For the The total number of class samples, For the The cluster center of a class sample is the mean vector of the feature vectors of similar samples.

[0112] The coverage radius is defined as the minimum hypersphere radius that contains 98% of similar samples, and the calculation formula is:

[0113] (19);

[0114] in, Represents the lower bound, that is, finding the minimum that meets the conditions ; It contains The minimum hypersphere radius of 98% of similar samples in the class, used to quantify the compactness of the distribution within the class; represents the candidate value of the hypersphere radius, covering the At least 98% of the samples in the class; For the The total number of class samples; It is an indicator function, which takes the value of 1 when the condition is met, and 0 otherwise.

[0115] Calculate the bi-norm distance between cluster centers of different categories:

[0116] (20);

[0117] in, 、 Respectively represent Class and The cluster center of the class sample, It is used to measure the degree of separation between any two types of samples in the feature space. The larger the distance, the stronger the inter-class separability.

[0118] By analyzing the measured data, it was found that the distance between bird flocks and precipitation and cloud targets was significantly higher than the distance between precipitation and cloud targets, proving that bird flock targets are significantly separable from meteorological targets in the feature space.

[0119] The overall process of a weather radar bird flock recognition and target difference analysis method in this application is as follows: Figure 4 As shown, its beneficial effects are:

[0120] (1) Based on the measured data of meteorological radar, the multi-dimensional feature analysis of bird flock targets, precipitation targets and cloud targets is carried out, including six features such as reflectivity, spectral width characteristics, and differential phase characteristics, making the results more credible;

[0121] (2) To extract the multidimensional characteristic information of bird flock targets, a series of refined analysis steps were performed to extract the characteristic data of migratory bird targets, laying the foundation for the characteristic analysis of bird flock targets and meteorological targets;

[0122] (3) The study is not only based on measured data, but also through statistical analysis and mathematical modeling, revealing the significant statistical differences in key characteristic parameters such as reflectivity and correlation coefficient in distinguishing bird flock targets from meteorological targets. It applies traditional statistical methods and clustering methods to new fields and provides a new technical path for multi-target recognition of meteorological radar data.

[0123] The present application also provides an application scenario, which applies the above-mentioned weather radar bird flock identification and target difference analysis method. Specifically: The weather radar bird flock identification and target difference analysis method provided in this embodiment can be applied in the weather radar bird flock target identification scenario. The weather radar bird flock target identification scenario includes a bird flock target identification link and a bird flock activity distribution map display link; the bird flock target identification link is used to identify bird flock targets from the weather radar raw data and generate a bird flock activity distribution map. The bird flock activity distribution map display link is used to display the bird flock activity distribution map; the weather radar bird flock identification and target difference analysis method provided in this embodiment belongs to the bird flock target identification link.

[0124] Based on the same inventive concept, the present application also provides a system for implementing the aforementioned method for weather radar bird flock identification and target difference analysis. The solution provided by this system is similar to the solution described in the aforementioned method. Therefore, the specific limitations of the one or more video tag processing device embodiments provided below can be found in the limitations of the weather radar bird flock identification and target difference analysis method described above and will not be repeated here.

[0125] The system includes:

[0126] The data acquisition module is used to obtain the original data of the weather radar and perform preprocessing operations on the original data of the weather radar to obtain the preprocessed weather radar data.

[0127] The bird flock target recognition module is used to: based on the first dynamic threshold method and the fuzzy logic classification method, perform a primary screening on the preprocessed meteorological radar data to obtain an initial bird flock target unit; the first dynamic threshold method is a preset first threshold of the multidimensional feature; the initial bird flock target unit is multiple; based on the spatiotemporal correlation analysis operation constraint condition, perform a secondary screening on the initial bird flock target unit to obtain the bird flock targets in the initial bird flock target unit; based on the initial bird flock target unit and the bird flock targets in the initial bird flock target unit, generate a bird flock activity distribution map.

[0128] The bird flock target feature extraction module is used to: extract multidimensional features from the pre-processed meteorological radar data based on the bird flock activity distribution map to obtain multidimensional feature information of the bird flock target; and construct a multidimensional feature matrix of the bird flock target based on the multidimensional feature information of the bird flock target.

[0129] like Figure 5 As shown, the target difference analysis module is used to: obtain the multidimensional feature matrix of the bird flock target and the multidimensional feature matrix of the meteorological target; use principal component analysis, Fisher linear discriminant method and clustering algorithm to process the multidimensional feature matrix of the bird flock target and the multidimensional feature matrix of the meteorological target to obtain the target separability verification result; the target separability verification result is used to quantify the separability of the bird flock target and the meteorological target.

[0130] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store processed data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for weather radar bird flock recognition and target difference analysis is implemented.

[0131] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0132] In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0133] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0134] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0135] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0136] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0137] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0138] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0139] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for bird flock identification and target difference analysis using weather radar, characterized in that: The weather radar bird flock identification and target difference analysis method includes: Acquiring raw weather radar data and performing a preprocessing operation on the raw weather radar data to obtain preprocessed weather radar data; Based on a first dynamic threshold method and a fuzzy logic classification method, the preprocessed weather radar data is screened to obtain an initial bird flock target unit; the first dynamic threshold method is a first threshold of a preset multidimensional feature; the initial bird flock target unit is multiple; Based on the spatiotemporal correlation analysis operation constraint condition, the initial bird flock target unit is secondary screened to obtain the bird flock targets in the initial bird flock target unit; generating a bird flock activity distribution map based on the initial bird flock target unit and the bird flock targets in the initial bird flock target unit; Based on the bird flock activity distribution map, multi-dimensional feature extraction is performed on the pre-processed weather radar data to obtain multi-dimensional feature information of the bird flock target; Constructing a bird flock target multidimensional feature matrix based on the bird flock target multidimensional feature information; Obtaining a multi-dimensional feature matrix of meteorological targets; The multidimensional feature matrix of the bird flock target and the multidimensional feature matrix of the meteorological target are processed using principal component analysis, Fisher linear discriminant method and clustering algorithm to obtain target separability verification results; the target separability verification results are used to quantify the separability of the bird flock target and the meteorological target.

2. The method for weather radar bird flock identification and target difference analysis according to claim 1, characterized in that: Performing a preprocessing operation on the raw weather radar data to obtain preprocessed weather radar data specifically includes: A second dynamic threshold method is used to remove ground clutter in the raw data of the weather radar to obtain a removal result; the second dynamic threshold method is a second threshold of a preset multi-dimensional feature; The elimination results are subjected to geographic coordinate transformation and orthogonal decomposition to obtain preprocessed weather radar data.

3. The method for weather radar bird flock identification and target difference analysis according to claim 1, characterized in that: Based on the first dynamic threshold method and the fuzzy logic classification method, the pre-processed weather radar data is screened to obtain the initial bird flock target unit, specifically including: Based on a first dynamic threshold method, non-bird flock target units in the preprocessed weather radar data are excluded to obtain potential bird flock target units; Calculating the target probability value of the potential bird flock target unit based on the fuzzy logic classification method to obtain the target probability value corresponding to the potential bird flock target unit; Based on the target probability values corresponding to the potential bird flock target units, the potential bird flock target units are screened to obtain initial bird flock target units.

4. The method for weather radar bird flock identification and target difference analysis according to claim 1, characterized in that: Based on the spatiotemporal correlation analysis operation constraint condition, the initial bird flock target unit is subjected to secondary screening to obtain the bird flock targets in the initial bird flock target unit, specifically including: Based on a preset neighborhood density, the initial bird flock target unit is processed to obtain a neighborhood density screening result; Based on a preset height dimension, the neighborhood density screening result is processed to obtain a height dimension screening result; Based on the preset spatiotemporal continuity constraints, the height dimension screening results are processed to obtain spatiotemporal correlation analysis results.

5. The method for weather radar bird flock identification and target difference analysis according to claim 1, characterized in that: The multi-dimensional features include reflectivity, radial velocity, spectral width, differential reflectivity, differential phase and correlation coefficient.

6. The method for weather radar bird flock identification and target difference analysis according to claim 3, characterized in that: Based on the fuzzy logic classification method, the target probability value of the potential bird flock target unit is calculated to obtain the target probability value corresponding to the potential bird flock target unit. The specific process is as follows: According to the preset trapezoidal probability function Calculate the target probability value; in, For the Target probability value of class target; is the weight matrix element, indicating the Class target characteristics In the The importance of class target recognition; For the Class target characteristics Corresponding to The probability of the target class.

7. The method for weather radar bird flock identification and target difference analysis according to claim 1, characterized in that: The multidimensional feature matrix of the bird flock target and the multidimensional feature matrix of the meteorological target are processed using principal component analysis, Fisher linear discriminant method and clustering algorithm to obtain target separability verification results, which specifically include: Processing the bird flock target multidimensional feature matrix and the meteorological target multidimensional feature matrix using principal component analysis to obtain a first dimensionality reduction feature space; The multidimensional feature matrix of the bird flock target and the multidimensional feature matrix of the meteorological target are processed by using Fisher linear discriminant method to obtain a second dimensionality reduction feature space; The first dimensionality reduction feature space and the second dimensionality reduction feature space are processed using a clustering algorithm to obtain a target separability verification result.

8. The method for weather radar bird flock identification and target difference analysis according to claim 1, characterized in that: The target separability verification results include cluster centers, coverage radius, and the bi-norm distance between cluster centers of different categories.

Images