A method and system for identifying bird echoes using China's S-band dual-polarization weather radar

Abstract

The present invention discloses a method and system for identifying bird echoes by China's S-band dual-polarization weather radar, the method comprising: pre-processing the characteristics of each observation parameter of my country's S-band dual-polarization weather radar base data; establishing the membership function and weight parameters of each parameter of the weather radar to form a fuzzy logic model; and realizing the accurate distinction between bird flocks and precipitation data by statistically determining the aggregated value of the membership function of the bird flock. Based on the four observation parameters and three derivatives of the dual-polarization radar, the present invention analyzes the probability density distribution function characteristics of the bird flock and precipitation data, constructs a fuzzy logic algorithm process and weight parameters suitable for my country's S-band dual-polarization radar to identify bird flocks, improves the accuracy of distinguishing bird flocks from precipitation echoes, and verifies the effectiveness of the method through individual case verification under different weather conditions such as birds leaving the nest, precipitation and clear sky, providing a new technical path for future weather radar biological echo identification and ecological monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of weather radar, and more particularly to a method and system for identifying bird echoes by a Chinese S-band dual-polarization weather radar. Background Art

[0002] Monitoring bird activities can not only provide a scientific basis for ecological protection and biodiversity, but also provide important data support for climate change research. By analyzing the migration routes of birds, the dynamic changes in habitats, and the health status of populations, we can understand the migration patterns and ecological needs of migratory birds, and provide a scientific basis for protecting migratory birds. At the same time, monitoring bird activities is also one of the important means to study the impact of climate change on organisms, providing important data support for how birds adapt to climate change and the impact of climate change on bird population dynamics. In addition, bird activities pose a potential threat to aviation safety, especially in areas around airports, where the risk of collisions between birds and aircraft is extremely high. Monitoring bird activities through radar can effectively reduce the occurrence of bird strikes and improve aviation safety.

[0003] In the fields of biology and ecological protection, a large number of studies have used insect radar to monitor bird activities. However, due to the small coverage of insect radar, the monitoring capacity is severely limited and cannot meet the needs of monitoring large-scale bird activities. At present, my country has built the world's largest weather radar observation network, with 546 weather radars deployed nationwide. Therefore, using weather radar data to identify bird activities has become a new technical monitoring method and has broad application prospects in meteorology, ecology and aviation safety.

[0004] Traditional methods for identifying bird flocks based on weather radar mainly rely on a single parameter (such as horizontal reflectivity factor ), this method has great limitations in distinguishing between organisms and precipitation. With the development of dual-polarization radar technology, more and more studies have shown that the differential reflectivity measured by dual-polarization radar , correlation coefficient and differential phase Polarization parameters can describe the shape, directionality and physical properties of target objects, and improve the ability to identify non-spherical particles (such as birds, insects, hail, etc.). For example, bird flocks are often shown as correlation coefficients in radar data. Low, and differential phase and differential reflectivity This provides a new technical means for accurately identifying bird flocks.

[0005] However, it was found in the application process that the commercialization time of my country's dual polarization radar is relatively short, the instrument performance varies greatly, and the The factors have greater deviations than those of foreign radars. The signal-to-noise ratio is a major factor affecting the accuracy of the identification results. This will cause a large error in the accuracy of the identification results when using fuzzy logic algorithms to identify bird flocks. Therefore, how to effectively use my country's dual-polarization radar data to identify bird flocks still lacks multi-parameter correction processing, comprehensive analysis, and comprehensive application of fuzzy logic algorithms. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a method and system for identifying bird echoes by China's S-band dual-polarization weather radar, so as to solve the problem of the dual-linear polarization radar in my country. The factors have greater deviations than those of foreign radars. The signal-to-noise ratio has a significant impact on the characteristics of the elements, resulting in unclear differences in the characteristics of bird flocks and precipitation in the polarization parameters, which makes it difficult to accurately distinguish bird flocks from meteorological echoes such as precipitation and clear sky using fuzzy logic technology.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] A method for identifying bird echoes by a Chinese S-band dual-polarization weather radar comprises the following steps:

[0009] Step P1: Preprocess the characteristics of each observation parameter of my country's S-band dual-polarization weather radar basic data, including threshold check, quality control, polarization correction, and derivative calculation, to further increase the differences between bird flocks and precipitation, clear sky clutter and other observation data characteristics;

[0010] Step P2: Based on the artificially confirmed weather radar data of real bird flock observations, the probability density distribution characteristics of bird flocks and precipitation observation data are statistically analyzed, the membership functions and weight parameters of the weather radar parameters are established, and a new fuzzy logic model for bird flock identification is formed;

[0011] Step P3: Through comprehensive judgment, the comprehensive judgment of the aggregation value of the bird flock membership function is statistically analyzed, and the optimal threshold of the aggregation value is screened to further achieve accurate distinction between the bird flock and precipitation data.

[0012] The above method is applicable to the identification of bird echoes by the S-band dual-polarization weather radar in China. In step P1, according to the element parameters observed by the S-band dual-polarization weather radar, four direct observation parameters are used and three derivatives are calculated for identification. The direct observation parameters are the horizontal reflectivity factor , correlation coefficient , differential reflectivity and differential phase ; The derivatives are the standard deviation of the differential reflectivity , standard deviation of differential phase as well as ; In the preprocessing stage, the four processing steps of threshold check, quality control, polarization correction and derivative calculation are carried out in sequence, and the execution order cannot be changed; according to the characteristics of the parameters, different parameters need to participate in different preprocessing steps respectively, as follows:

[0013] Step P1-1: Horizontal reflectivity factor and correlation coefficient Perform threshold checks;

[0014] Step P1-2: Correlation coefficient , differential reflectivity and differential phase Perform quality control of spatial filtering and smoothing;

[0015] Step P1-3: Correlation coefficient , differential phase Perform polarization correction processing;

[0016] Step P1-4: Standard deviation of reflectivity difference for three derived parameters , standard deviation of differential phase as well as Calculate the derivative amount.

[0017] Four direct observation parameters and three calculated diffraction quantities in the dual polarization radar are selected as input parameter data. The four direct observation parameters are preprocessed according to their respective characteristics, which can increase the difference characteristics of each parameter when observing precipitation and bird flocks.

[0018] The above method for identifying bird echoes by China's S-band dual-polarization weather radar retains the horizontal reflectivity factor in step P1-1. The threshold is between 15 dBZ and 50 dBZ to remove the interference of clear sky echo; the correlation coefficient is retained. The data with a value less than 0.9 is selected to eliminate the interference of the rainy area data.

[0019] The above method for identifying bird echoes by China's S-band dual-polarization weather radar, in step P1-2, performs the following checks based on the results processed by P1-1:

[0020] Spatial filtering: Correlation coefficient , differential reflectivity , differential phase Parameters, using a local window-based method, with the target pixel as the center, considering both the radial and distance libraries, defining a 5×5 window area, and calculating the number of valid values ​​of all pixels in the window. If there are more than 10 pixels in the window with valid values, the data of the center point is retained, otherwise it is set as a missing value;

[0021] Smoothing: Correlation coefficient , differential reflectivity The two parameters are smoothed by taking the average value of seven consecutive pixels on the same radial direction; , the median smoothing method of 7 consecutive pixels on the same radial direction is used.

[0022] The above method for identifying bird echoes by China's S-band dual-polarization weather radar, in step P1-3, performs the following checks based on the results processed by P1-2:

[0023] For differential phase , polarization correction is divided into two parts: initial value correction and self-consistent correction:

[0024] Initial value correction: Based on the historical precipitation data of a single station, using the close-range echo method to determine The initial value is used to correct the observation data of each pixel point within the radar coverage area by subtracting the initial value;

[0025] Self-consistent correction: According to the statistical precipitation conditions in previous studies and The self-consistent relationship is calculated by formula (1) changes in rain areas and subtracting that change from the observations to highlight differences in bird populations and precipitation; (1);

[0026] In formula (1), Indicates i The horizontal reflectivity factor of each radar pixel, in mm -6 m -3 ; Indicates i The differential reflectivity of radar pixels is in degrees. When the absolute value is greater than 4.0dB, take ; dr Indicates radial distance in km; Indicates i The self-consistent correction value of the differential phase of each radar pixel is in degrees. When it is less than 0, take , if this library has no valid data, =0;

[0027] Get the corrected differential phase ,Right now: ;

[0028] For the correlation coefficient , the signal-to-noise ratio is used for correction to obtain the corrected correlation coefficient , as shown in formula (2):

[0029] (2)

[0030] In formula (2), is the signal-to-noise ratio.

[0031] The above method for identifying bird echoes by China's S-band dual-polarization weather radar, in step P1-4, performs the following checks based on the results processed by P1-3:

[0032] Standard deviation of differential reflectance and the standard deviation of the differential phase The standard deviation formula is used for calculation, as shown in formula (3):

[0033] (3)

[0034] In formula (3), represents the radar parameter, that is and , and Respectively represent the range of 5°×1.25KM Mean and squared mean;

[0035] The calculation of is obtained by using formula (4):

[0036] (4);

[0037] In formula (4), ;

[0038] Since precipitation data has the characteristic of positive skew distribution in statistics, we need to take the logarithm of the SDR result and then multiply it by 10 to get , in order to better distinguish precipitation from bird flock data; as shown in formula (5):

[0039] (5).

[0040] The above method for identifying bird echoes by China's S-band dual-polarization weather radar includes, in step P2:

[0041] Step P2-1: Constructing the membership function: Use piecewise function to construct the membership function of each parameter in the fuzzy logic algorithm. The output value of the membership function is between 0 and 1. The larger the value, the higher the possibility that the parameter is a flock of birds within the function interval. For areas where the probability density of flocks is very large and the probability density of precipitation is very small, the ordinate of the node parameter of the membership function is set to 1, otherwise it is set to 0; obtain the node parameters of the membership function of all parameters;

[0042] Step P2-2: According to the node parameters of the membership function of each parameter, a function with a basic form of a trapezoid is used to realize the fuzzification of the parameter to obtain a fuzzy logic model.

[0043] The above method, which is applicable to the identification of bird echoes by China's S-band dual-polarization weather radar, determines the weights in the fuzzy logic model algorithm: the weight of each parameter is proportional to the distinguishing ability of the probability density distribution function of the parameter, that is, the greater the difference between the distribution of the bird flock and the rain area, the better the identification effect, and the greater the weight should be, otherwise the smaller the weight;

[0044] The weights of each parameter are: differential reflectivity The weight is 5%, the correlation coefficient The weight is 20%, the horizontal reflectivity factor The weight is 15%, differential phase The weight of the differential reflectivity is 25%, the weight of the SDR is 5%, and the standard deviation of the differential reflectivity is The weight is 10%, the standard deviation of the differential phase The weight is 20%.

[0045] The above method is applicable to the identification of bird echoes by China's S-band dual-polarization weather radar. In step P3, the membership function and weight of the input parameter are weighted averaged to calculate the aggregation value Q of the radar data point by point. The calculation method is as follows:

[0046] (7);

[0047] In formula (7), Indicates the number of membership functions used in the fuzzy logic algorithm, Indicates i The weight of the parameter, Indicates i The membership function of the parameters; Q represents the probability that the observation result at this point is identified as a flock of birds;

[0048] The fuzzy logic algorithm is used to make a comprehensive judgment on the aggregate value Q and the confidence level: the aggregate value Q is sorted from large to small, and the concept of quantile is used, with special attention paid to the 95% quantile of the confidence level, and the threshold of the aggregate value Q is quantitatively given as 0.82;

[0049] That is, when the Q calculated from the radar data of a certain point is greater than 0.82, the point is determined to be bird flock data; otherwise, the point is determined to be non-bird flock data.

[0050] The probability density of bird flocks under clear sky conditions was compared with that under precipitation conditions, and a single membership function was constructed after comprehensive analysis, which simplified the calculation process of traditional fuzzy logic algorithms using multiple membership functions and reduced the complexity of the algorithm.

[0051] A bird echo recognition system suitable for China's S-band dual-polarization weather radar, including: data processing module: pre-processing the characteristics of each observation parameter of my country's S-band dual-polarization weather radar base data, including four steps of threshold check, quality control, polarization correction, and derivative calculation, to obtain the horizontal reflectivity factors of four direct observation parameters after pre-processing , correlation coefficient , differential reflectivity and differential phase and the standard deviation of the three derivatives , standard deviation of differential phase as well as ;

[0052] Function building module: used to build membership function and weight distribution function based on fuzzy logic algorithm and radar echo types;

[0053] Classification module: Count the aggregate values ​​of the bird flock membership function, and filter the optimal threshold of the aggregate value to further achieve accurate distinction between bird flocks and precipitation data.

[0054] The technical solution of the present invention achieves the following beneficial technical effects:

[0055] 1. Based on the fuzzy logic algorithm, this method comprehensively analyzes and processes the multi-parameter data of the S-band dual-polarization radar. By comparing and analyzing the probability density functions of bird echoes and precipitation echoes, a new single membership function is constructed, and a new bird echo recognition algorithm combining threshold judgment with fuzzy logic is proposed.

[0056] 2. This method aims at the characteristics of dual polarization radar data in my country and the insufficiency of multi-parameter comprehensive analysis in bird flock identification. Based on the multi-parameter data of dual polarization radar (such as , , and etc.), optimize the fuzzy logic algorithm process and parameters, construct the probability density distribution function of bird flock and precipitation data suitable for my country's S-band dual-polarization radar, improve the accuracy of distinguishing bird flocks from precipitation echoes, and verify the effectiveness of this method through individual cases under different weather conditions such as birds leaving the nest, precipitation and clear sky, providing a new technical path for future weather radar biological echo identification and ecological monitoring.

[0057] 3. This application selects 4 direct observation parameters and 3 diffraction calculation parameters in dual polarization radar as input data. The 4 direct observation parameters are preprocessed according to their respective characteristics, which can increase the difference characteristics of each parameter when observing precipitation and bird flocks.

[0058] right and Perform threshold check. By setting specific thresholds, the interference of clear sky echo and rain area data can be removed respectively. , differential reflectivity and differential phase Perform spatial filtering and smoothing quality control. Initial value correction and self-consistent correction were performed to highlight the differences between bird flocks and precipitation, and the correlation coefficient The signal-to-noise ratio is corrected to reduce the interference of clutter signals; for the three calculated derivatives, the standard deviation of the differential reflectivity is used , standard deviation of differential phase It reflects the degree of change in the amount of polarization and uses SDR to distinguish particles (such as raindrops) from asymmetric particles (such as birds) to increase recognition accuracy.

[0059] Through the analysis of the probability density distribution of the seven parameters before and after preprocessing, the distribution range of the parameter data after preprocessing becomes narrower, the data distribution becomes more concentrated, and the overlapping area of ​​the bird flock data and the precipitation data distribution area becomes smaller, especially , , This shows that although the two types of echoes overlap in some parameters, they can be effectively distinguished through multi-parameter joint analysis.

[0060] 4. From the recognition results of various parameters before and after preprocessing, it can be clearly seen that when using the preprocessed parameters for recognition, due to the significant distribution differences between the parameters in the bird flock area and other areas, the recognized bird flock data is large and the ring structure is obvious, that is, the recognition effect is good.

[0061] Under clear sky conditions, bird flock echoes and clear sky echoes can be accurately distinguished, especially using pre-processed parameters, the circular or semi-ring structure of the bird flock is more obvious and the recognition effect is better.

[0062] Under precipitation conditions, the use of preprocessed parameters can completely eliminate the interference of precipitation echoes and achieve better recognition effects.

[0063] 5. This application compares the probability density of bird flocks under clear sky conditions with the probability density under precipitation conditions, and after comprehensive analysis, constructs a single membership function. Through the comprehensive judgment method of fuzzy logic algorithm aggregation value and confidence level analysis and observation, the threshold of the aggregation value Q is quantitatively given as 0.82. This simplifies the calculation process of traditional fuzzy logic algorithm using multiple membership functions and reduces the complexity of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a flow chart of the bird flock recognition algorithm of the S-band dual-polarization weather radar based on fuzzy logic of the present invention;

[0065] Figure 2 Aggregate value and confidence probability distribution curve;

[0066] Figure 3a For example 1, the parameters use the recognition results before preprocessing. Figure 3b The recognition results of each parameter after preprocessing are shown in Example 1;

[0067] Figure 4a For example 2, the parameters use the recognition results before preprocessing. Figure 4b The recognition results of each parameter after preprocessing are used for the parameters of case 2;

[0068] Figure 5a For example 3, the parameters use the recognition results before preprocessing. Figure 5b The recognition results of each parameter after preprocessing for case 3 are shown. DETAILED DESCRIPTION

[0069] 1. Fuzzy Logic Algorithm

[0070] 1.1. This invention proposes an algorithm combining threshold judgment and fuzzy logic to identify bird flock data. The algorithm is divided into three parts. The specific process is as follows: Figure 1 shown.

[0071] First, before using fuzzy logic to identify bird flocks, the single-station radar base data observations are preprocessed, including threshold checking, quality control, polarization correction and derivative calculation, which increases the difference characteristics of various parameters when observing precipitation and bird flocks.

[0072] Secondly, through the artificially confirmed real bird flock observation data, the probability density distribution characteristics of bird flock and precipitation observation data were statistically analyzed, the membership function and weight parameters of each parameter were re-established, and a new fuzzy logic model was formed.

[0073] Finally, the bird flock and precipitation data were accurately distinguished through comprehensive judgment of the aggregated values ​​of the bird flock membership function.

[0074] 1.2 Parameter preprocessing

[0075] In the recognition algorithm, the input radar direct observations have four parameters, namely: 1) horizontal reflectivity factor (Unit: dBZ): reflects the target object's ability to reflect radar waves. 2) Correlation coefficient : Indicates the degree of correlation between horizontal and vertical polarization signals. 3) Differential reflectivity (Unit: dB): Indicates the difference in reflectivity of the target object in the horizontal and vertical polarization directions. 4) Differential Phase (Unit: °): represents the phase difference between horizontal and vertical polarization signals. The input calculation diffraction parameters have three parameters, namely, the standard deviation of differential reflectivity , standard deviation of differential phase as well as .

[0076] In the preprocessing stage, four processing steps are performed in sequence: threshold check, quality control, polarization correction, and derivative calculation. The execution order cannot be changed. Different parameters need to participate in different preprocessing steps according to the characteristics of the parameters.

[0077] 1.2.1 Threshold Check

[0078] First, the radar data characteristics of bird flocks, precipitation and clear days were analyzed. and The threshold check was performed. The horizontal reflectivity factor is an important parameter used in radar detection to measure the total scattering ability of scatterers within the target volume. Its value can reflect the scattering characteristics of the target object. This algorithm retains the horizontal reflectivity factor The threshold is between 15 dBZ and 50 dBZ to remove the interference of clear sky echo. It can characterize the consistency of the shape, size and scattering characteristics of the scatterers in the echo volume, and is an effective indicator for identifying different types of target objects. This algorithm retains the correlation coefficient The data with a value less than 0.9 is selected to eliminate the interference of the rainy area data.

[0079] 1.2.2 Quality Control

[0080] Polarization parameters , , Quality control was performed. Spatial filtering and smoothing were performed only based on the original observation base data before identification.

[0081] Spatial filtering uses a local window-based method, with the target pixel as the center, considering both the radial and distance libraries, defining a 5×5 window area, and calculating the number of valid values ​​of all pixels in the window. If there are more than 10 pixels in the window with valid values, the data of the center point is retained, otherwise it is set to a missing value.

[0082] To further enhance the smoothness and consistency of the data, the polarization parameters , , Smoothing in the radial direction is performed. , Parameters, the average value of 7 consecutive pixels on the same radial direction (with the current pixel as the center) is used for smoothing; Taking into account its phase characteristics, the median smoothing method of 7 consecutive pixels on the same radial direction is used to avoid the phase shift phenomenon caused by the average value calculation, ensuring that the smoothed data can more realistically reflect the characteristics of change.

[0083] 1.2.3 Polarization Correction

[0084] In order to further divide the differences between bird flocks and precipitation data, , Carry out correction processing.

[0085] Data correction is divided into two parts: initial value correction and self-consistent correction:

[0086] Initial value correction: Based on the historical precipitation data of a single station, using the close-range echo method to determine The initial value is used to correct the observation data of each pixel point within the radar coverage area by subtracting the initial value;

[0087] The self-consistent correction is based on the statistical precipitation conditions in previous studies. and The statistical self-consistency relationship of is calculated by formula (1) The change in rain area is then subtracted from the observations to highlight differences in bird populations and precipitation.

[0088] (1); In formula (1), Indicates i The horizontal reflectivity factor of each radar pixel, in mm -6 m -3 ; Need to be converted from dBZ; Indicates i The differential reflectivity of radar pixels is in degrees. When the absolute value is greater than 4.0dB, take ; dr Indicates radial distance in km; Indicates i The self-consistent correction value of the differential phase of each radar pixel is in degrees. When it is less than 0, take , if this library has no valid data, =0;

[0089] Get the corrected differential phase ,Right now: ;

[0090] For the correlation coefficient , the signal-to-noise ratio is used for correction to obtain the corrected correlation coefficient , as shown in formula (2):

[0091] (2)

[0092] In formula (2), is the signal-to-noise ratio.

[0093] 1.2.4. Derivative Quantity Calculation

[0094] The algorithm introduces three derivatives to participate in the fuzzy logic identification, namely, the standard deviation of the differential reflectivity , standard deviation of differential phase as well as , where the standard deviation can reflect the degree of change in polarization. SDR can be used to distinguish particles (such as raindrops) from asymmetric particles (such as birds). The specific algorithm is as follows, as shown in formula (3):

[0095] (3)

[0096] In formula (3), represents the radar parameter, that is and , and Respectively represent the range of 5°×1.25km Mean and squared mean;

[0097] The calculation of is obtained by using formula (4):

[0098] (4);

[0099] In formula (4), ;

[0100] It should be noted that since precipitation data has the characteristic of positive skew distribution in statistics, the SDR result should be logarithmized and then multiplied by 10 to obtain , in order to better distinguish precipitation from bird flock data.

[0101] (5).

[0102] In Section 1.2, four directly observed parameters and three diffraction calculated parameters were preprocessed, including: and Perform threshold check. By setting specific thresholds, the interference of clear sky echo and rain area data can be removed respectively. , differential reflectivity and differential phase Perform spatial filtering and smoothing quality control. Initial value correction and self-consistent correction were performed to highlight the differences between bird flocks and precipitation, and the correlation coefficient The signal-to-noise ratio is corrected to reduce the interference of clutter signals; for the three calculated derivatives, the standard deviation of the differential reflectivity is used , standard deviation of differential phase Reflecting the degree of change in polarization, SDR is used to distinguish particles (such as raindrops) from asymmetric particles (such as birds) to increase recognition accuracy. In this way, preprocessing is performed according to the characteristics of each parameter, which can increase the difference characteristics of each parameter when observing precipitation and bird flocks.

[0103] 1.3 Probability Density Distribution

[0104] In order to verify the effect of preprocessing of each parameter, the manually labeled bird flock data in case 1 and the large-scale precipitation data in case 2 are used to analyze the parameters of the bird flock and precipitation data. , , , , , , ,After analyzing the probability density distribution before and after preprocessing, the parameter data distribution interval becomes narrower after threshold checking, quality control and correction, the data distribution becomes more concentrated, and the overlapping area of ​​the bird flock data and the precipitation data distribution area becomes smaller, especially , , This shows that although the two types of echoes overlap in some parameters, they can be effectively distinguished through multi-parameter joint analysis.

[0105] 1.4. Constructing membership functions

[0106] Based on the analysis of the probability density function, the piecewise function is used to construct the membership function of each parameter of the fuzzy logic algorithm. The output value of the membership function ranges from 0 to 1. The larger the value, the higher the possibility that the parameter is a bird flock within the function interval. For areas with very large probability density of bird flocks and very small probability density of precipitation, the ordinate of the node parameter of the membership function is set to 1, otherwise it is set to 0.

[0107] by For example, after preprocessing, the bird flock The values ​​of are mainly concentrated below 0.6, while the rain area The probability that the value of is below 0.6 is very small and close to 0, so it is determined that: When it is less than 0.6, the characteristics of the bird flock are significant, and the membership function of the correlation coefficient is The node parameter is set to 1; similarly, when hour, It has no flock characteristics at all, and the node parameter of the membership function of this point is equal to 0. All the membership function node parameters of . According to this method, the membership function node parameters of all parameters can be obtained. See Table 1.

[0108] Table 1 Membership function node parameters

[0109]

[0110] According to the node parameters of the membership function of each parameter, the fuzzification of the parameter is realized by using a function with a basic form of trapezoid. In the process of constructing the membership function, not only the probability density distribution of the bird flock is considered, but also the difference in the probability density distribution of the bird flock and the rain area is considered. Therefore, a single membership function can be used to identify the data.

[0111] 1.5. Determination of weights

[0112] The weight of each parameter is proportional to the distinguishing ability of the probability density distribution function of the parameter, that is, the greater the difference between the distribution of the bird flock and the rain area, the better the recognition effect, and the greater the weight should be, and vice versa. Based on this principle, the weights of each parameter in the fuzzy logic algorithm of this embodiment are given as follows:

[0113] Table 2 Weights of fuzzy logic algorithms

[0114]

[0115] 1.6 Comprehensive judgment

[0116] Finally, the membership function and weight of the input parameters are used to calculate the aggregate value of the radar data point by point through weighted averaging. , calculated as follows:

[0117] (7)

[0118] In formula (7), Indicates the number of membership functions used in the fuzzy logic algorithm (in this study ), represents the weight of the i-th parameter, represents the membership function of the ith parameter. Q represents the probability that the observation result at this point is identified as a flock of birds.

[0119] like Figure 2 As shown in the figure, the aggregation value and confidence of the fuzzy logic algorithm are analyzed and observed, and the aggregation values ​​are sorted from large to small. The concept of quantile is used, and special attention is paid to the 95% quantile of the confidence. Under the assumption of normal distribution, the confidence corresponding to 95% of the aggregation values ​​is not lower than this value, and only 5% of the aggregation values ​​have a lower confidence. Therefore, the threshold of the aggregation value Q is quantitatively given as 0.82. That is, when the Q calculated by the radar data of a certain point is greater than 0.82, the point is judged as bird flock data, otherwise, the point is judged as non-bird flock data.

[0120] 2. Inspection

[0121] The Liaohe Delta wetland near the Yingkou radar station is an important transit station and habitat for migrating waterfowl from East Asia to Australia. There are a large number of bird migration activities in spring and autumn every year. Three case data of the Yingkou radar station were selected for verification, and the specific information is shown in Table 3. All cases were provided by the staff of the Yingkou radar station and contain information on manually marked bird flock activities. Among them, case 1 is the radar observing bird flock activities under clear sky conditions, and the bird flock location information is manually marked after on-site verification by the station staff. Case 2 is the observation of weather radar under precipitation conditions to verify the omission of the algorithm. Case 3 is the simultaneous observation of bird flocks and precipitation within the observation range, which is used to verify the classification and recognition of the algorithm when precipitation and bird flocks exist at the same time. The time of the bird flock cases was around sunrise in the early morning, and the radar data used were all 0.5° elevation angle data.

[0122] Table 3 Yingkou weather radar case data

[0123]

[0124] The radar data of case 1 is identified using fuzzy logic algorithm. Figure 3a and Figure 3b The following are the recognition results of the data before and after preprocessing using this algorithm. When the aggregation value of the identified point is greater than the threshold, the point is judged as a flock of birds, represented by blue. When the aggregation value of the identified point is less than the threshold, it is judged as non-flock of birds, represented by yellow (the same below). Figure 3a and Figure 3bIt can be seen from the figure that the bird flock data identified in the northwest direction of the radar are in a circular shape when the pre-processing parameters are used for identification, and the circular structure is very obvious; more bird flock data are also identified in the northeast and southwest directions of the radar, but no circular structure is formed, so it is impossible to directly and accurately judge whether the identification result is accurate; a small number of point-like identification results appearing in the southeast direction of the radar have no regular distribution and do not conform to the distribution characteristics of the bird flock, which is the result of the misjudgment of the ground clutter as a bird flock. When the pre-processed parameters are used for identification, the distribution difference between the corrected parameters in the bird flock area and other areas increases, making the identified bird flock data more: the circular structure of the identification result in the northwest direction of the radar is more significant; the identification result in the northeast direction of the radar also presents circular and semi-circular features, and the identification result in the southwest direction of the radar has a smaller range of disc-shaped features, which can be judged that there are also bird flock activities in this direction; the identification result in the southeast direction of the radar has a point-like distribution feature, which does not conform to the distribution characteristics of the bird flock. After manual on-site verification, it is determined that the ground clutter is misjudged as a bird flock (that is, the basis for misjudgment and missed judgment is the conclusion drawn from manual on-site verification). It can be seen that the fuzzy logic algorithm proposed in this study can accurately distinguish between bird flock echoes and clear sky echoes under clear sky conditions, especially using the preprocessed parameters, the recognition effect is better, and it is suitable for use in my country's dual polarization radar.

[0125] The radar data of case 2 is identified using fuzzy logic algorithm. Figure 4a and Figure 4b The recognition results of the data before and after preprocessing are shown in the figure. Figure 4a It can be seen that using the parameters before preprocessing for identification, there are misjudged bird flock data within 70km above the radar. Figure 4b It can be seen that when using the preprocessed parameters for identification, the precipitation data information can be completely eliminated. It can be seen that the fuzzy logic algorithm proposed in this study can well eliminate precipitation data.

[0126] The radar data of case 3 is identified using fuzzy logic algorithm. Figure 5a and Figure 5b The recognition results of the data before and after preprocessing are shown in the figure. Figure 5a It can be seen that when the parameters before preprocessing are used for identification, the bird flock data identified in the northwest direction within the range of 150km of the radar present multiple ring structures, which overlap with each other, and the characteristics of the bird flock are significant; multiple semicircular ring structures appear in the northeast and southwest directions of the radar, and it can be judged that there are also bird flocks in this location; there are no ring or semicircular ring structures in the southeast direction of the radar, and the distribution is relatively scattered, which is consistent with the characteristics of ground clutter. The algorithm can remove most of the precipitation echoes far away from the radar, but there are cases of missed rejection. According to statistics, the missed rejection rate is 0.26% ( Figure 5a There is an error in the identification of precipitation areas in the distant radar). Figure 5b It can be seen that when the pre-processed parameters are used for identification, the identification of bird flocks is similar to that of the parameters before pre-processing, but the precipitation echoes far away from the radar can be completely eliminated. It can be seen that the fuzzy logic algorithm proposed in this study can well identify the echo information of precipitation conditions, especially using the pre-processed parameters, which can completely eliminate precipitation data.

[0127] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the claims of this patent application.

Claims

1. A method for identifying bird echoes using a Chinese S-band dual-polarization weather radar, characterized in that: The steps include: Step P1: Preprocess the characteristics of each observation parameter of my country's S-band dual-polarization weather radar basic data, including threshold check, quality control, polarization correction, and derivative calculation, to increase the difference between the characteristics of bird flocks and precipitation and clear sky clutter observation data; Step P2: Based on the artificially confirmed weather radar data of real bird flock observations, the probability density distribution characteristics of bird flocks and precipitation observation data are statistically analyzed, the membership functions and weight parameters of the weather radar parameters are established, and a fuzzy logic model for bird flock identification is formed; Step P3: Through comprehensive judgment, the aggregate value of the bird flock membership function is counted, and the optimal threshold of the aggregate value is screened to achieve accurate distinction between the bird flock and precipitation data; In step P1, according to the element parameters observed by the S-band dual-polarization weather radar, four direct observation parameters are used and three derivatives are calculated for identification. The direct observation parameters are the horizontal reflectivity factor Z H , correlation coefficient ρ HV , differential reflectivity Z DR and differential phase ; The derived quantities are the standard deviation of the differential reflectivity σ( Z DR ), standard deviation of differential phase and SDR; In the preprocessing stage, four processing steps are carried out in sequence: threshold check, quality control, polarization correction, and derivative calculation. According to the characteristics of the parameters, different parameters need to participate in different preprocessing steps, as follows: Step P1-1: Horizontal reflectivity factor Z H and correlation coefficient ρ HV Perform threshold checks; Step P1-2: Correlation coefficient ρ HV , differential reflectivity Z DR and differential phase Perform quality control of spatial filtering and smoothing; Step P1-3: Correlation coefficient ρ HV , differential phase Perform polarization correction processing; Step P1-4: The standard deviation σ of the differential reflectivity of the three derivative parameters ( Z DR ), standard deviation of differential phase and SDR for derivative volume calculation; Z H and ρ HV Perform threshold check and set specific thresholds to remove interference from clear sky echoes and rainy area data respectively; ρ HV , differential reflectivity Z DR and differential phase Perform spatial filtering and smoothing quality control; perform differential phase Initial value correction and self-consistent correction were performed to highlight the differences between bird flocks and precipitation. ρ HV The signal-to-noise ratio is corrected to reduce the interference of clutter signals; for the three calculated derivatives, the standard deviation of the differential reflectivity σ( Z DR ), standard deviation of differential phase It reflects the degree of change in polarization and uses SDR to distinguish particles from asymmetric particles, increasing recognition accuracy.

2. The method for identifying bird echoes by a Chinese S-band dual-polarization weather radar according to claim 1, characterized in that: In step P1-1, the horizontal reflectivity factor Z is retained H The threshold is between 15 dBZ and 50 dBZ to eliminate the interference of clear-air echo; Keep the correlation coefficient ρ HV The data with a value less than 0.9 is selected to eliminate the interference of the rainy area data.

3. The method for identifying bird echoes by a Chinese S-band dual-polarization weather radar according to claim 2, characterized in that: In step P1-2, the following checks are performed based on the results of P1-1: Spatial filtering: Correlation coefficient ρ HV , differential reflectivity Z DR , differential phase Parameters, using a local window-based method, with the target pixel as the center, considering both the radial and distance libraries, defining a 5×5 window area, and calculating the number of valid values ​​of all pixels in the window. If there are more than 10 pixels in the window with valid values, the data of the center point is retained, otherwise it is set as a missing value; Smoothing: Correlation coefficient ρ HV , differential reflectivity Z DR The two parameters are smoothed by taking the average value of seven consecutive pixels on the same radial direction; , the median smoothing method of 7 consecutive pixels on the same radial direction is used.

4. The method for identifying bird echoes by a Chinese S-band dual-polarization weather radar according to claim 3, characterized in that: In step P1-3, the following checks are performed based on the results of P1-2: For differential phase , polarization correction is divided into two parts: initial value correction and self-consistent correction: Initial value correction: Based on the historical precipitation data of a single station, using the close-range echo method to determine The initial value is used to correct the observation data of each pixel point within the radar coverage area by subtracting the initial value; Self-consistent correction: According to the statistical precipitation conditions in previous studies Z DR and Z H The self-consistent relationship is calculated by formula (1) changes in rain areas and subtracting that change from the observations to highlight differences in bird populations and precipitation; (1); In formula (1), Z Hi Indicates i The horizontal reflectivity factor of each radar pixel, in mm -6 m -3 ; Z DRi Indicates i The differential reflectivity of radar pixels is in degrees. Z DRi When the absolute value is greater than 4.0dB, take ; dr Indicates radial distance in km; Indicates i The self-consistent correction value of the differential phase of each radar pixel is in degrees. When it is less than 0, take If this library has no valid data, ; Get the corrected differential phase ,Right now: ; For the correlation coefficient ρ HV , the signal-to-noise ratio is used for correction to obtain the corrected correlation coefficient , as shown in formula (2): (2) In formula (2), is the signal-to-noise ratio.

5. The method for identifying bird echoes by a Chinese S-band dual-polarization weather radar according to claim 4, characterized in that: In step P1-4, the following checks are performed based on the results of P1-3: The standard deviation of the differential reflectivity σ( Z DR ) and the standard deviation of the differential phase The standard deviation formula is used for calculation, as shown in formula (3): (3) In formula (3), x represents the radar parameter, that is Z DR and , and Respectively represent the range of 5°×1.25km x Mean and squared mean; The SDR is calculated using formula (4): (4)； In formula (4), ; Since precipitation data has the characteristic of positive skewed distribution in statistics, we need to take the logarithm of the SDR result and multiply it by 10 to obtain SDR', so as to better distinguish precipitation from bird flock data; as shown in formula (5): （5）。 6. The method for identifying bird echoes by a Chinese S-band dual-polarization weather radar according to claim 5, characterized in that: In step P2, Step P2-1: Constructing the membership function: Use piecewise function to construct the membership function of each parameter in the fuzzy logic algorithm, and the output value of the membership function is 0 to 1: for areas with very high probability density of bird flocks and very low probability density of precipitation, set the ordinate of the node parameter of the membership function to 1, otherwise set it to 0; obtain the node parameters of the membership function of all parameters; Step P2-2: According to the node parameters of the membership function of each parameter, a function with a basic form of a trapezoid is used to realize the fuzzification of the parameter to obtain a fuzzy logic model.

7. The method for identifying bird echoes by a Chinese S-band dual-polarization weather radar according to claim 6, characterized in that: Determine the weights in the fuzzy logic algorithm: The weight of each parameter is proportional to the distinguishing ability of the probability density distribution function of the parameter, that is, the greater the difference between the distribution of the bird flock and the rain area, the better the recognition effect, and the greater the weight should be, otherwise the weight should be smaller; The weights of each parameter are: differential reflectivity Z DR The weight is 5%, the correlation coefficient ρ HV The weight is 20%, the horizontal reflectivity factor Z H The weight is 15%, differential phase The weight of is 25%, the weight of SDR is 5%, and the standard deviation of differential reflectivity σ( Z DR ) with a weight of 10% and the standard deviation of the differential phase The weight is 20%.

8. The method for identifying bird echoes by a Chinese S-band dual-polarization weather radar according to claim 7, characterized in that: In step P3, the membership function and weight of the input parameters are used to calculate the aggregate value Q of the radar data point by point through weighted averaging. The calculation method is as follows: (7)； In formula (7), k Indicates the number of membership functions used in the fuzzy logic algorithm, w i Indicates i The weight of the parameter, f ( x i ) indicates the i The membership function of the parameters; Q represents the probability that the observation result at this point is identified as a flock of birds; The fuzzy logic algorithm is used to make a comprehensive judgment on the aggregate value Q and the confidence level: the aggregate value Q is sorted from large to small, and the concept of quantile is used, with special attention paid to the 95% quantile of the confidence level, and the threshold of the aggregate value Q is quantitatively given as 0.82; That is, when the Q calculated from the radar data of a certain point is greater than 0.82, the point is determined to be bird flock data; otherwise, the point is determined to be non-bird flock data.

9. A bird echo recognition system suitable for China's S-band dual-polarization weather radar, characterized in that: include: Data processing module: preprocess the characteristics of each observation parameter of my country's S-band dual-polarization weather radar basic data, including threshold check, quality control, polarization correction, and derivative calculation, and obtain the horizontal reflectivity factor Z of the four direct observation parameters after preprocessing H , correlation coefficient ρ HV , differential reflectivity Z DR and differential phase and the standard deviations of the three derivatives σ( Z DR ), standard deviation of differential phase and SDR; In the preprocessing stage, four processing steps are carried out in sequence: threshold check, quality control, polarization correction, and derivative calculation. According to the characteristics of the parameters, different parameters need to participate in different preprocessing steps, as follows: Step P1-1: Horizontal reflectivity factor Z H and correlation coefficient ρ HV Perform threshold checks; Step P1-2: Correlation coefficient ρ HV , differential reflectivity Z DR and differential phase Perform quality control of spatial filtering and smoothing; Step P1-3: Correlation coefficient ρ HV , differential phase Perform polarization correction processing; Step P1-4: The standard deviation σ of the differential reflectivity of the three derivative parameters ( Z DR ), standard deviation of differential phase and SDR for derivative volume calculation; Z H and ρ HV Perform threshold check and set specific thresholds to remove interference from clear sky echoes and rainy area data respectively; ρ HV , differential reflectivity Z DR and differential phase Perform spatial filtering and smoothing quality control; perform differential phase Initial value correction and self-consistent correction were performed to highlight the differences between bird flocks and precipitation. ρ HV The signal-to-noise ratio is corrected to reduce the interference of clutter signals; for the three calculated derivatives, the standard deviation of the differential reflectivity σ( Z DR ), standard deviation of differential phase It reflects the degree of change in polarization, and uses SDR to distinguish particles from asymmetric particles, increasing recognition accuracy; Function construction module: Statistical probability density distribution characteristics of bird flock and precipitation observation data, establish membership functions and weight parameters of various weather radar parameters, and form a fuzzy logic model for bird flock identification; Classification module: Count the aggregate values ​​of the bird flock membership function, and filter the optimal threshold of the aggregate value to further achieve accurate distinction between bird flocks and precipitation data.

Images