阅读说明：本技术 基于朴素贝叶斯分类器的双偏振雷达二次回波识别方法 (Dual-polarization radar secondary echo identification method based on naive Bayes classifier ) 是由 邵世卿 侯小宇 刘淑 徐喜东 于 2021-08-30 设计创作，主要内容包括：本发明涉及一种基于朴素贝叶斯分类器的双偏振雷达二次回波识别方法,包括使用双偏振雷达I/Q数据,计算当前距离库用于二次回波识别的四个判据：PDE-(h)、PDE-(v)、PDE-(hv)以及SQI；根据这四个判据的数值,查找预先统计好的各判据CPDF,获取指定数值在不同分类下的概率,计算该距离库回波数据属于一次回波的概率P-(f)和属于二次回波的概率P-(s)；当P-(s)＞P-(f)时,判定该距离库回波为二次回波,否则判定为一次回波；继续进行下一个距离库的识别。该方法提高了在雷达数据层面的二次回波识别精度,对双偏振雷达参量估算质量的提升有重要意义。(The invention relates to a secondary echo recognition method of a dual-polarization radar based on a naive Bayes classifier, which comprises the following steps of calculating four criteria of a current distance library for secondary echo recognition by using I/Q data of the dual-polarization radar: PDE h 、PDE v 、PDE hv And SQI; according to the values of the four criteria, searching each pre-counted criterion CPDF, acquiring the probability of the designated value under different classifications, and calculating the probability P that the echo data of the distance library belongs to the primary echo f And probability P of being a secondary echo s (ii) a When P is present s ＞P f Judging the echo of the distance library as a secondary echo if the distance library is normal, otherwise, judging the echo as a primary echo; continuing to proceed to the next distanceAnd (5) identifying the off-library. The method improves the secondary echo identification precision on the radar data level, and has important significance for improving the estimation quality of the dual-polarization radar parameters.)

1. A dual-polarization radar secondary echo recognition method based on a naive Bayes classifier is characterized by comprising the following steps: the method specifically comprises the following steps:

step S1: forming a distance library from the I/Q data acquired by the dual-polarization radar, and calculating by adopting the distance library to acquire four judgment bases for secondary echo identification, wherein the four judgment bases are respectively a phase distribution evaluation parameter and a signal quality factor, and the phase distribution evaluation parameter comprises a horizontal channel parameter of the dual-polarization radar, a vertical channel parameter of the dual-polarization radar and a parameter between a horizontal channel and a vertical channel;

step S2: according to the four obtained judgment bases, searching each pre-statistically formed criterion, namely a conditional probability density function, obtaining the primary echo probability and the secondary echo probability of the designated numerical value under different classifications, judging whether the echo data in the distance library belongs to the primary echo probability or the secondary echo probability, and defining the primary echo probability as P_fThe probability of secondary echo is P_s；

Step S3: if P_s＞P_fJudging the echo data in the distance library to be secondary echoes, otherwise judging the echo data in the distance library to be primary echoes;

step S4: and repeating the process, and continuing to identify the next distance library.

2. The naive bayes classifier based dual polarization radar secondary echo identification method according to claim 1, wherein: in step S1The phase distribution evaluation parameter is defined as PDE, and the horizontal channel parameter of the dual polarization radar, the vertical channel parameter of the dual polarization radar, and the parameter between the horizontal and vertical channels are respectively defined as PDE_h、PDE_vAnd a PDE_hv；

The phase distribution evaluation parameter PDE is calculated by the formula

In equation (1), when PDE calculation is required_hWhen it is necessary to calculate PDE, take x (n) ═ h (n)/h (n +1)_vWhen x (n) is equal to v (n)/v (n +1), PDE calculation is required_hvWhen x (N) is equal to h (N)/v (N), N represents a distance bank, h represents horizontal channel data, v represents vertical channel data, and N represents the number of relevant pulses.

3. The naive bayes classifier based dual polarization radar secondary echo identification method according to claim 2, wherein: and when the PDE value is small, judging the echo data to be a secondary echo.

4. The naive bayes classifier based dual polarization radar secondary echo identification method according to claim 1, wherein: in step S1, the signal quality factor is defined as SQI, which is calculated by the formula

In the formula (2), R₁Representing a first order autocorrelation function, i.e. ACF, R₀Represents a zero order ACF, wherein the calculation formula of the ACF is

In formula (3), M represents the number of relevant pulses, n represents the order, a represents the conjugate, V_h，vThe voltage value of a horizontal channel or a vertical channel is represented, M represents the serial number of the pulse, and M is more than or equal to 1 and less than or equal to M.

5. The naive bayes classifier based dual polarization radar secondary echo identification method according to claim 1, wherein: in step S2, the conditional probability density function is defined as CPDF, which requires several samples for statistics, i.e. at least one month of observation data, and only uses the horizontal channel parameter of the dual-polarization radar or the vertical channel parameter of the dual-polarization radar or the parameter between the horizontal and vertical channels as parameters.

6. The naive Bayes classifier based dual polarization radar secondary echo identification method of claim 5, wherein: in step S2, P_fThe calculation method of (2) assumes the current distance library as a primary echo, then searches the probability corresponding to the calculated value of each criterion under the condition of the primary echo in the CPDF, and finally multiplies the probability values of several criteria.

7. The naive Bayes classifier based dual polarization radar secondary echo identification method of claim 5, wherein: in step S2, P_sThe calculation method of (2) assumes that the current distance library is a quadratic echo, then searches the probability corresponding to the calculated value of each criterion under the condition of the quadratic echo in the CPDF, and finally multiplies the probability values of several criteria.

8. The naive bayes classifier based dual polarization radar secondary echo identification method according to claim 1, wherein: in actual operation, a combined conditional probability density function of a horizontal channel parameter of the dual-polarization radar and a vertical channel parameter of the dual-polarization radar is adopted, and the calculation method is that the horizontal channel parameter and the vertical channel parameter are simultaneously used as parameters, and the distribution condition is differentiated from a sample.

Technical Field

The invention relates to a secondary echo recognition method of a dual-polarization radar based on a naive Bayes classifier, belongs to the field of radar signal processing research, and particularly relates to a secondary echo recognition method of a weather radar.

Background

Secondary echo is an important factor affecting the quality of weather radar data, and is also a noise source relative to primary echo. For a particular weather radar, at a given pulse repetition Period (PRT) T_sWorking down, maximum unambiguous distance r_aAnd maximum unambiguous velocity v_aIs then determined, where r_a＝cT_s/2，v_a＝λ/(4T_s) C is the speed of light and λ is the radar wavelength. The product of the maximum unambiguous distance and the maximum unambiguous velocity is a constant, r_av_aC λ/8, so at a given λ, r is increased_aWill result in v_aSmaller and vice versa, also known as Doppler dilemma (Bringi and Chandrasekar,2001, Doviak and2006, Zhang Pegchang et al 2001). When a pulse detects an object beyond the maximum unambiguous distance r_aIn time, echoes may form in subsequent pulses, resulting in the appearance of secondary echoes (collectively referred to herein as secondary echoes for multiple echoes).

Methods for resolving secondary echoes are numerous and can be broadly divided into two broad categories, the first being methods for phase encoding the transmit pulse (Laird,1981, Zrinc and Mahapatra,1985, sachidanda and zrnc, 1986b, sachidanda and zrnc, 1999, Torres,2008, Bharadwaj and Chandrasekar,2007), and the second being radar modes of operation using multiple PRTs or frequencies (PRFs) (sirans et al, 1976, Zrinc and Mahapatra,1985, sachidanda and zrnc, 2003, Torres et al, 2004, Cho, 2005). The basic principle of the phase encoding method is that when a radar receiver synchronizes only primary echoes, if the phase of a transmitted pulse is randomly encoded, the received secondary echoes become random signals and are represented as noise. Sachidana et al (1986b) proposeA systematic phase encoding method (SZ encoding) is disclosed which performs better than the random phase method in separating the secondary echoes. On this basis, a modified SZ encoding method SZ (8/64) is proposed and applied to WSR-88D radar (Torres, 2008). For the multiple PRT/PRF method, briefly (taking dual PRT as an example), a long repetition period T is used₁To solve the distance ambiguity problem and then use a short repetition period T₂To address speed ambiguity issues, such as batch mode (Torres et al, 2004) for WSR-88D radar. Theoretically, this multi-frequency approach can achieve infinite unambiguous range and unambiguous velocity, but the method is limited by such factors as limited radar sampling, secondary echo strength, and weather echo correlation time (Cao et al, 2012 b).

In recent years, a secondary echo recognition method based on radar data (base data or I/Q data) has been proposed. Cao et al (2012b) propose a secondary echo identification method widely applicable to magnetron radars, which is based on the principle of a random Phase method, and can effectively identify secondary echoes (weak secondary echoes, weak secondary echoes and strong secondary echoes) affected to different degrees by using a PDE (Phase Distribution Evaluation) and a fuzzy logic method. Park et al (2016) use Z_DR、Ф_DP、ρ_hv、σ_vAnd (5) the standard deviation and the mean value of the equal parameters are used for identifying secondary echoes of the dual-polarization radar based on a fuzzy logic method. The data-level-based identification method has the advantages that the existing radar hardware is used as a basis, hardware equipment does not need to be additionally arranged, the algorithm is more flexible, and the identification effect is poor.

In China, research has also been conducted for many years in the aspect of identification and processing of secondary echoes. A new phase coding sequence and a corresponding frequency domain processing method are provided by Zhu Hua et al (2002), and simulation shows that the method has better ambiguity resolution performance and is easy to process in real time. Dong gem et al (2006) studied and discussed methods for dual frequency de-blurring, phase encoding SZ (8/64), and batch processing. Pan Xinmin and the like (2010) discuss a method for removing ambiguity of a new-generation weather radar in China, and the method comprises two aspects of distance removing ambiguity and speed removing ambiguity. Liu Sheng Feng (2014) and the like use random phase encoding to resolve range ambiguity for X-band Doppler weather radar. And (2017) performing ambiguity resolution application of a random phase encoding method on the S-band weather radar. In general, domestic research work on secondary echoes is also carried out in the aspects of phase encoding and multi-PRT/PRF.

The method uses phase coding or a multiple PRT/PRF secondary echo processing method, and the realization process needs the support of radar hardware and a signal processing algorithm. However, the identification of secondary echoes in the radar data plane is mostly based on a fuzzy logic algorithm, and certain engineering experience is needed when determining the member functions of each criterion. If a method based on a naive Bayesian Classifier (SBC) is used, the recognition result depends on a Conditional Probability Density Function (CPDF) of a criterion, and the statistics can be directly carried out through a sample, so that the operation method is more deterministic.

Disclosure of Invention

The invention provides a secondary echo recognition method of a dual-polarization radar based on a naive Bayes classifier, which is suitable for intermediate-frequency coherent radars and full-coherent radars of which transmitted pulses adopt phase coding and has better recognition effect.

The technical scheme adopted by the invention for solving the technical problems is as follows:

a dual-polarization radar secondary echo identification method based on a naive Bayes classifier specifically comprises the following steps:

step S1: forming a distance library from the I/Q data acquired by the dual-polarization radar, and calculating by adopting the distance library to acquire four judgment bases for secondary echo identification, wherein the four judgment bases are respectively a phase distribution evaluation parameter and a signal quality factor, and the phase distribution evaluation parameter comprises a horizontal channel parameter of the dual-polarization radar, a vertical channel parameter of the dual-polarization radar and a parameter between a horizontal channel and a vertical channel;

step S2: according to the four obtained judgment bases, searching each pre-statistically formed criterion, namely a conditional probability density function, and obtaining one of the specified values under different classificationsJudging whether the echo data in the distance library belongs to the primary echo probability or the secondary echo probability, and defining the primary echo probability as P_fThe probability of secondary echo is P_s；

Step S3: if P_s＞P_fJudging the echo data in the distance library to be secondary echoes, otherwise judging the echo data in the distance library to be primary echoes;

step S4: repeating the above process, and continuing to identify the next distance library;

as a further preferable aspect of the present invention, in step S1, the phase distribution evaluating parameter is defined as PDE, and the horizontal channel parameter of the dual polarization radar, the vertical channel parameter of the dual polarization radar, and the parameter between the horizontal and vertical channels are respectively defined as PDE_h、PDE_vAnd a PDE_hv；

The phase distribution evaluation parameter PDE is calculated by the formula

In equation (1), when PDE calculation is required_hWhen it is necessary to calculate PDE, take x (n) ═ h (n)/h (n +1)_vWhen x (n) is equal to v (n)/v (n +1), PDE calculation is required_hvWhen the pulse width is larger than the preset value, x (N) ═ h (N)/v (N) is taken, N represents a distance library, h represents horizontal channel data, v represents vertical channel data, and N represents the number of relevant pulses;

as a further preferred aspect of the present invention, when the PDE value is large, the echo data is determined to be a primary echo, and when the PDE value is small, the echo data is determined to be a secondary echo;

as a further preferred embodiment of the present invention, in step S1, the signal quality factor is defined as SQI, and the calculation formula is

In the formula (2), R₁Representing a first order autocorrelation function, i.e. ACF，R₀Represents a zero order ACF, wherein the calculation formula of the ACF is

In formula (3), M represents the number of relevant pulses, n represents the order, a represents the conjugate, V_h，vThe voltage value of a horizontal channel or a vertical channel is represented, M represents the serial number of the pulse, and M is more than or equal to 1 and less than or equal to M;

as a further preferred embodiment of the present invention, in step S2, the conditional probability density function is defined as CPDF, which requires several samples for statistics, i.e. at least one month of observation data, and only uses the horizontal channel parameter of the dual-polarization radar or the vertical channel parameter of the dual-polarization radar or the parameter between the horizontal and vertical channels as parameters;

as a further preferred aspect of the present invention, in step S2, P_fThe calculation method of (1) assumes the current distance library as a primary echo, then searches the probability corresponding to the calculation value of each criterion under the condition of the primary echo in the CPDF, and finally multiplies the probability values of several criteria;

as a further preferred aspect of the present invention, in step S2, P_sThe calculation method of (1) assumes the current distance library as a quadratic echo, then searches the probability of the calculation value of each criterion under the condition of the quadratic echo in the CPDF, and finally multiplies the probability values of several criteria;

as a further preferred aspect of the present invention, in the actual operation, a joint conditional probability density function of the horizontal channel parameter of the dual polarization radar and the vertical channel parameter of the dual polarization radar is used, and the calculation method is to take the horizontal channel parameter and the vertical channel parameter as parameters at the same time, and to count the distribution condition from the sample.

Through the technical scheme, compared with the prior art, the invention has the following beneficial effects:

1. the invention provides a secondary echo recognition method of a dual-polarization radar based on a naive Bayes classifier, which is a method based on mathematical statistics, does not depend on more experience factors, has more determinacy during operation and better recognition effect;

2. the bi-polarization radar secondary echo recognition method based on the naive Bayes classifier, provided by the invention, is based on radar I/Q data only, and has the advantages of high automation degree, convenience, easiness and high applicability.

Drawings

The invention is further illustrated with reference to the following figures and examples.

FIG. 1 is a schematic algorithm flow diagram of a preferred embodiment provided by the present invention;

FIGS. 2 a-2 d are conditional probability density functions for two types of echoes (primary echo and secondary echo) for the criteria of the preferred embodiment statistics provided by the present invention;

fig. 3 a-3 d are comparison of the recognition results of two echoes of a single PPI scan according to the preferred embodiment of the present invention.

Detailed Description

The present invention will now be described in further detail with reference to the accompanying drawings. In the description of the present application, it is to be understood that the terms "left side", "right side", "upper part", "lower part", etc., indicate orientations or positional relationships based on those shown in the drawings, and are only for convenience of describing the present invention and simplifying the description, but do not indicate or imply that the referred device or element must have a specific orientation, be constructed in a specific orientation, and be operated, and that "first", "second", etc., do not represent an important degree of the component parts, and thus are not to be construed as limiting the present invention. The specific dimensions used in the present example are only for illustrating the technical solution and do not limit the scope of protection of the present invention.

For the secondary echo caused by the meteorological echo, the secondary echo is still the meteorological echo essentially, when the radar works in the single repetition frequency mode, the conventional parameter and the polarization parameter of the radar have normal estimated values, and only because the positions of the secondary echo and the primary echo are different, the secondary echo and the primary echo have difference in data continuity; thus, Park et al (2016) use Z_DR、Ф_DP、ρ_hv、σ_vStandard deviation and mean of equal parameters, and fuzzy logic method based on the methodAnd identifying secondary echoes. If the echoes near the radar are all secondary echoes, then the method using echo continuity and texture is not necessarily feasible, as the secondary echoes themselves may also be relatively continuous. For the primary echo, the numerical range of the parameter estimation is greatly overlapped with the secondary echo, so that the secondary echo cannot be well confirmed by only using one or more radar parameters.

Based on the above, the application aims to provide a secondary echo recognition method for a dual-polarization radar based on a naive Bayes classifier, which is based on a mathematical method of the naive Bayes classifier, firstly, according to a data sample prepared in advance, the prior probability of each recognition criterion is counted, and then, the prior probability is used for calculating the posterior probability, so that a classification result of a target is obtained; therefore, compared with a series of traditional identification methods based on fuzzy logic and the like, the identification method provided by the application has the advantages that the operation is more deterministic and the identification effect is better when the member functions are determined to depend on more empirical factors.

As shown in fig. 1, the specific steps include step S1: and forming a distance library by using the I/Q data acquired by the dual-polarization radar, and calculating by using the distance library to acquire four judgment bases for secondary echo identification, wherein the four judgment bases are respectively a phase distribution evaluation parameter and a signal quality factor, and the phase distribution evaluation parameter comprises a horizontal channel parameter of the dual-polarization radar, a vertical channel parameter of the dual-polarization radar and a parameter between a horizontal channel and a vertical channel.

Step S2: according to the four obtained judgment bases, searching each pre-statistically formed criterion, namely a conditional probability density function, obtaining the primary echo probability and the secondary echo probability of the designated numerical value under different classifications, judging whether the echo data in the distance library belongs to the primary echo probability or the secondary echo probability, and defining the primary echo probability as P_fThe probability of secondary echo is P_s；

The conditional probability density function is defined as CPDF, a plurality of samples are needed during statistics, namely observation data of at least one month are needed, and only the horizontal channel parameter of the dual-polarization radar or the vertical channel parameter of the dual-polarization radar or the parameter between the horizontal channel and the vertical channel is adopted as a parameter; meanwhile, the CPDF has pertinence, and different radars need to count probability density distribution functions conforming to the characteristics of the radars; when the same radar is used in different areas (for example, the areas from the areas of the Yangtze river and the Huaihe river to the areas with the triangular beads), the CPDF is required to be updated due to the change of the precipitation characteristics so as to obtain a better secondary echo recognition effect.

P_fThe calculation method of (1) assumes the current distance library as a primary echo, then searches the probability corresponding to the calculation value of each criterion under the condition of the primary echo in the CPDF, and finally multiplies the probability values of several criteria;

P_sthe calculation method of (2) assumes that the current distance library is a quadratic echo, then searches the probability corresponding to the calculated value of each criterion under the condition of the quadratic echo in the CPDF, and finally multiplies the probability values of several criteria.

Step S3: for a primary echo, the phases of adjacent distance bins are relatively close, and the PDE value is relatively large; for the secondary echo, the phase difference of adjacent distance libraries is large, and the PDE value is small; therefore, the classification with the greater probability value is selected as the recognition result of the current distance database, specifically, if P_s＞P_fAnd judging the echo data in the distance library to be secondary echoes, otherwise, judging the echo data in the distance library to be primary echoes.

Step S4: and repeating the process, and continuing to identify the next distance library.

In step S1, the phase distribution evaluation parameter, i.e., PDE, was set forth by Cao et al (2012b) in the method for identifying secondary echoes of its magnetron radar, and in this application, the horizontal channel parameter of the dual polarization radar, the vertical channel parameter of the dual polarization radar, and the parameter between the horizontal and vertical channels are defined as PDE, respectively_h、PDE_vAnd a PDE_hvThe formula of PDE is

In equation (1), when PDE calculation is required_hWhen x (n) is equal to h (n)/h (n +1), the compound is represented byNeed to calculate the PDE_vWhen x (n) is equal to v (n)/v (n +1), PDE calculation is required_hvWhen x (N) ═ h (N)/v (N), N represents a distance library, h represents horizontal channel data, v represents vertical channel data, N represents the number of relevant pulses, and hv represents the combination of the horizontal channel and the vertical channel.

In addition, in step S1, a signal quality factor, SQI, is provided, which can well express the data quality of radar returns, and thus also serves as one of the criteria of the secondary recognition algorithm, in combination with PDE_h、PDE_vAnd a PDE_hvThe secondary echoes are identified together, and the calculation formula is

In the formula (2), R₁Representing a first order autocorrelation function, i.e. ACF, R₀Represents a zero order ACF, wherein the calculation formula of the ACF is

In formula (3), M represents the number of relevant pulses, n represents the order, a represents the conjugate, V_h，vThe voltage value of a horizontal channel or a vertical channel is represented, M represents the serial number of the pulse, M is more than or equal to 1 and less than or equal to M, and h and v represent the horizontal channel or the vertical channel.

Example (b):

the application provides an embodiment as verification, taking data acquired by an NJU-CPOL radar as an example, the data is acquired by a C-band dual-polarization radar (NJU-CPOL) of Nanjing university in 6-7 months in 2014, the place is Anhui Changfeng, and a conditional probability density function of the data is shown in figures 2 a-2 d, wherein First-Trip represents a primary echo, Second-Trip represents a secondary echo, and the CPDF statistics in the figures uses observation data of two months.

As can be seen from fig. 2a and 2b, the first and second echoes are selected as joint conditional probability density functions, since the PDE_hAnd a PDE_vHave similarities withAnd the combined conditional probability density function is used in actual operation, the calculation method of the combined conditional probability density function is different from that of the conditional probability density function, the calculation method of the combined conditional probability density function takes the horizontal channel parameter and the vertical channel parameter as parameters at the same time, the distribution condition is unified from a sample, the combined probability density function is used, the target characteristics are identified from the perspective of binary random variables, more information can be provided than that provided by the unitary random variables, and the accuracy of the identification method can be further improved.

Then respectively calculating P with the location of Anhui Changfeng_fAnd P_sData are acquired by an NJU-CPOL radar at 37 minutes (UTC) at 12 days and 10 days in 7 months in 2014, data are shown in FIGS. 3a to 3d, which are comparison results of secondary echo identification results scanned by PPI in a primary precipitation process and discrimination results of a random phase method (a phase encoding method), a radar echo map containing secondary echoes is shown in FIG. 3a, a result after secondary echoes are removed is shown in FIG. 3b, a secondary echo identification result based on an SBC is shown in FIG. 3c, and an identification result of the random phase method is shown in FIG. 3d, and identification results are found to be basically the same after comparison of the two results.

In summary, the two-polarization radar secondary echo recognition method based on the naive Bayes classifier has a better recognition effect than a fuzzy logic method, is suitable for intermediate-frequency coherent radars and full-coherent radars of which transmitted pulses adopt phase coding, and is simpler and more efficient than a hardware signal processing method of phase coding.

It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

The meaning of "and/or" as used herein is intended to include both the individual components or both.

The term "connected" as used herein may mean either a direct connection between components or an indirect connection between components via other components.

In light of the foregoing description of the preferred embodiment of the present invention, many modifications and variations will be apparent to those skilled in the art without departing from the spirit and scope of the invention. The technical scope of the present invention is not limited to the content of the specification, and must be determined according to the scope of the claims.