Radar echo classification and identification method based on multi-factor parameter features

Abstract

This invention proposes a radar echo classification and identification method based on multi-factor parameter features. By calculating the feature factors for classifying radar echoes, the method determines the feature thresholds of the discriminant factor parameters for different types of radar echoes, extracts the discriminant factor variables for different radial range circles of the radar, and establishes a radar echo classification and discrimination model. This invention can automatically monitor and identify radar echoes of ground objects, clear skies, and precipitation, providing a scientific reference for the identification of different types of echoes. It automates and objectifies the work that previously required subjective analysis and identification of radar echo images by meteorological professionals, improving the accuracy and reliability of automatically identifying different types of echoes.

Description

Technical Field

[0001] This invention relates to the field of atmospheric science and technology, and in particular to a radar echo classification and identification method based on multi-factor parameter features. Background Technology

[0002] Weather radar echoes are categorized into meteorological and non-meteorological echoes. Non-meteorological echoes include singular echoes, ground feature echoes, and super-refractive echoes, while meteorological echoes include precipitation and non-precipitation echoes. The primary impact on radar data usage comes from non-meteorological echoes such as ground feature and super-refractive echoes, which can affect weather interpretation and even lead to misjudgments. Although the signal processors of new-generation weather radar systems have the ability to cancel ground feature clutter in echo signals, they cannot completely eliminate the influence of non-precipitation echoes. Further research is needed on radar echo classification and identification methods to improve radar data quality.

[0003] In recent years, research both domestically and internationally has focused on radar echo classification and recognition methods based on fuzzy logic, horizontal and vertical reflectivity factor structures, and neural networks.

[0004] In 2003, Kessinger proposed the "Radar Echo Classifier (REC)". This algorithm takes three basic radar data points (reflectivity factor, radial velocity, and spectral width) as input. Depending on the REC algorithm, different feature combinations are selected as input to obtain a series of features. These features are then processed using fuzzy logic techniques to generate three products for classifying radar echoes. Under certain conditions, this algorithm can effectively remove singular echoes and ground feature echoes, but its performance is not ideal for identifying mixed echoes of ground features and precipitation.

[0005] In 2004, Zhang et al. proposed a quality control algorithm based on horizontal and vertical reflectivity factor structures. This algorithm is based on the fact that precipitation echoes and non-precipitation echoes have different horizontal and vertical reflectivity factor structures. The algorithm is simple and efficient, and can effectively identify non-precipitation echoes such as ground clutter, bird flocks under clear skies, and insect echoes. However, it is difficult to distinguish between strange echoes at a distance and weakly stratiform cloud precipitation echoes.

[0006] In 2005, Lakshmanan et al. proposed a neural network method that integrates local mean values ​​such as echo intensity, velocity, and spectral width, as well as echo texture features and SPIN. The algorithm's advantage lies in its excellent performance when processing echo data containing various contaminants. However, its disadvantages include the large number of features required for computation, and the fact that not all features can be calculated for every point, resulting in long computation time and relatively low algorithm efficiency. Summary of the Invention

[0007] To address the shortcomings and deficiencies of existing technologies, this invention proposes a radar echo classification and identification method based on multi-factor parameter features. This method calculates and discriminates radar echo characteristic factors, statistically analyzes the characteristic thresholds of discriminant factor parameters for different types of radar echoes, determines the discriminant factor variables for different radial range circles of the radar, and establishes a radar echo classification and discrimination model. This invention can automatically monitor and identify radar echoes of ground objects, clear skies, and precipitation, providing a scientific reference for the identification of different types of echoes.

[0008] This method calculates characteristic factors for classifying radar echoes, determines the characteristic thresholds of discriminant factor parameters for different types of radar echoes, extracts discriminant factor variables for different radial range circles of the radar, and establishes a radar echo classification and discrimination model. This invention can automatically monitor and identify radar echoes of ground objects, clear skies, and precipitation, providing a scientific reference for the identification of different types of echoes. It automates and objectifies the work that previously required subjective analysis and identification of radar echo images by meteorological professionals, improving the accuracy and reliability of automatically identifying different types of echoes.

[0009] The present invention specifically adopts the following technical solution:

[0010] A radar echo classification and identification method based on multi-factor parameter features, characterized by the following steps:

[0011] Step S1: Acquire the base data detected by CINRAD / SA weather radar;

[0012] Step S2: Remove isolated noise interference from radar base data;

[0013] Step S3: Calculate the feature factors used to classify radar echoes:

[0014] For the radar base data after quality control in step S2, calculate radar echo characteristic factors including at least the echo standard deviation STDEV, horizontal reflectivity factor structure TDBZ, reflectivity factor horizontal texture, reflectivity factor vertical gradient VertGrad, echo inter-column variation SPIN, echo radial sign change SIGN, radial velocity region average MDVE, velocity variance SDVE, spectral width region average MDSW, and echo top height EchoTop.

[0015] Step S4: Based on statistical analysis of individual radar echo cases of different types, determine the threshold values ​​of the discriminant factor characteristic parameters for different types of radar echoes;

[0016] Step S5: Filter the discrimination factors for different radial range circles of the radar;

[0017] Step S6: Based on step S5, determine the influencing factors of the discrimination factors for each interval, and establish a radar echo classification and discrimination model.

[0018] Furthermore, in step S1, the CINRAD / SA weather radar base data is arranged from low to high elevation angle: 0.5-19.5 degrees, the elevation angle scan data of each layer is arranged according to radial scan lines 0-360 degrees, and each radial scan line is stored in polar coordinates according to a distance library of 1 km x 1 degree.

[0019] Furthermore, in step S2, the removal of isolated noise from radar base data is achieved by taking a selected echo point as the center and comparing the percentage of invalid reflectivity factor values ​​P within a 5×5 window with a threshold to remove isolated noise interference.

[0020] Furthermore, in step S3, the specific calculation methods for the characteristic factors used to classify radar echoes, including echo standard deviation (STDEV), horizontal reflectivity factor structure (TDBZ), reflectivity factor horizontal texture (Texture), reflectivity factor vertical gradient (VertGrad), echo inter-column variation (SPIN), echo radial sign change (SIGN), radial velocity region average (MDVE), velocity variance (SDVE), spectral width region average (MDSW), and echo top height (EchoTop), are as follows:

[0021] (1) Echo standard deviation ZSTDEV

[0022]

[0023] N: Represents the number of echo points in the template, X i : Echo intensity with subscript i The average value of the echo intensity within the template;

[0024] (2) Horizontal reflectivity factor structure TDBZ

[0025]

[0026] Nrays and Ngates: These represent the calculation ranges defined along the azimuth and radial directions, respectively. i.j This represents the value of the reflectivity factor;

[0027] (3) Reflectivity Factor Horizontal Texture

[0028]

[0029] (4) Inter-column variation of echo along the radial direction SPIN

[0030]

[0031]

[0032] Zthresh represents the threshold for changes in inter-column echo intensity;

[0033] (5) Intercolumn sign change of the echo along the radial direction (SIGN)

[0034]

[0035]

[0036] (6) Vertical gradient of reflectivity factor VertGrad

[0037]

[0038] Z up and Z low θ represents the echo values ​​of the corresponding echo point at the high elevation angle layer and the low elevation angle layer, respectively. up θ low These represent the elevation angle values ​​for the high-rise and low-rise buildings, respectively.

[0039] (7) Median Radial Velocity MDVE: Radial velocity value after median filtering of echo points;

[0040] (8) Velocity Variance SDVE

[0041]

[0042] X i.j This represents the radial velocity value of the echo points within the selected area. The median velocity within the selected area;

[0043] (9) Median velocity spectrum width MDSW: The velocity spectrum width after median filtering of the echo point;

[0044] (10) EchoTop

[0045]

[0046] Z t Echo threshold, 5dBZ, Z h and Z l H represents the echo intensity values ​​corresponding to the elevation angles of the upper and lower floors, respectively. h and H l These represent the echo point heights at the corresponding high and low elevation angles, respectively; the echo height H is obtained using the altimeter formula.

[0047]

[0048] Where h0 represents the radar antenna height, R represents the slant range between the target and the radar, α represents the elevation angle, and Rm is the equivalent Earth radius.

[0049] As a preferred embodiment, in step S4, the threshold values ​​for the discrimination factor characteristic parameters of different types of radar echoes are as follows:

[0050] Ground features and superrefractive echoes: The horizontal reflectivity factor structure (TDBZ) and horizontal texture (TEXTURE) are large, the SPIN value is relatively large, the vertical gradient of reflectivity factor (VertGrad) is less than 0 and the absolute value is very large, the median radial velocity (MDVE) is close to 0, the velocity variance (SDVE) and median velocity spectrum width (MDSW) are both very small, and the echo top height (EchoTop) is between 3 and 4.5 km.

[0051] Clear-sky echo: The echo intensity is relatively low, generally less than 15 dBZ; the echo standard deviation ZSTDEV is small, the SPIN value is small, and the echo top height EchoTop is generally less than 3 km.

[0052] Precipitation echoes: TDBZ and TEXTURE are generally less than 30, SPIN value is small, vertical gradient is closer to 0, and EchoTop is greater than 5 km.

[0053] Furthermore, in step S5, the discrimination factors for different radial range circles of the radar are as follows: in the range of 0-50km, R, ZSTDEV, TDBZ, TEXTURE, VertGrad, SPIN, MDVE, SDVE, and MDSW are used as discrimination factors; in the range of 50-150km, R, ZSTDEV, TDBZ, TEXTURE, VertGrad, SPIN, MDVE, SDVE, MDSW, and EchoTop are used as discrimination factors; and in the range of >150km, R, ZSTDEV, TDBZ, TEXTURE, VertGrad, SPIN, and EchoTop are used as discrimination factors.

[0054] Due to the influence of radar scanning mode and its own detection capability, different echo characteristic factors should be selected for different radial distances. The effective radius of radar radial velocity data is generally around 150 km, and the maximum detection radius is 230 km. Therefore, the statistical interval is divided into three parts: 0-50 km, 50-150 km, and greater than 150 km.

[0055] Furthermore, in step S6, the influencing factors of the discriminant factors for each interval are determined and a Logistic regression model is established.

[0056] The logistic regression model is mainly used to predict the probability of events influenced by multiple factors. It represents the occurrence of an event as Y = 0, and the non-occurrence of an event as Y = 1. E(Y) = p represents the probability of an event occurring, X1…X… m These are the predictor factors; then, the coefficients of the Logistic model are calculated based on the sample data.

[0057] ln(p / (1-p))=b0+b1X1+...b m X m +ε=b0+β'X+ε

[0058] The maximum likelihood method is typically used to estimate b0b1…b m That is, from:

[0059]

[0060] Solve for X, where n is the number of observations; k is the number of events; and X is the number of occurrences. i It is the observed value of the predictor factor when the event occurs, which is X. j Observed values ​​of the forecast factor when the event does not occur.

[0061] This invention and its preferred embodiments calculate and discriminate radar echo characteristic factors, statistically analyze the characteristic thresholds of discriminant factor parameters for different types of radar echoes, determine the discriminant factor variables for different radial range circles of the radar, and establish a radar echo classification and discrimination model. This invention can automatically monitor and identify radar echoes of ground objects, clear skies, and precipitation, providing a scientific reference for the identification of different types of echoes. Attached Figure Description

[0062] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0063] Figure 1 This is an overall flowchart of the method according to an embodiment of the present invention;

[0064] Figure 2 This is a technical roadmap of the method according to an embodiment of the present invention;

[0065] Figure 3 This describes the radar echo classification and recognition effect in this embodiment of the invention. Figure 1 ;

[0066] Figure 4 This describes the radar echo classification and recognition effect in this embodiment of the invention. Figure 2 . Detailed Implementation

[0067] In the following, specific embodiments of this application will be described in detail with reference to the accompanying drawings. Based on these detailed descriptions, those skilled in the art will be able to clearly understand and implement this application. Without departing from the principles of this application, features from various embodiments can be combined to obtain new implementations, or certain features from some embodiments can be substituted to obtain other preferred implementations.

[0068] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0069] To make the features and advantages of this patent more apparent and understandable, specific embodiments are provided below for detailed explanation:

[0070] like Figure 1 , Figure 2 As shown, the radar echo classification and recognition method based on multi-factor parameter features provided in this embodiment includes the following steps:

[0071] Step S1: Acquire basic data from CINRAD / SA weather radar.

[0072] The most significant feature of the CINRAD / SA weather radar is its volume scanning method based on a volume coverage model (VCP). It scans twice at each elevation angle in the lower layers, which is recorded as one elevation layer in the base data. Taking VCP21 mode as an example, the VCP elevation angles are: 0.5, 0.5, 1.5, 1.5, 2.4, 3.4, 4.3, 6.0, 9.9, 14.6, and 19.5, for a total of 11 PPI scans. The radar base data is arranged from low to high elevation angle (0.5-19.5 degrees), and the elevation scan data for each layer is stored in polar coordinates using radial scan lines from 0 to 360 degrees, with each radial scan line stored in a 1 km × 1 degree range library.

[0073] Step S2: Remove isolated noise interference from radar base data

[0074] Centered on a certain echo point, take a 5×5 window and calculate the percentage P of all detected data values ​​within the window that are invalid reflectivity factor values. If P is greater than a set threshold, then the point is determined to be isolated noise interference.

[0075] Step S3: Calculate the feature factors used to classify radar echoes.

[0076] For the radar base data after quality control in step S2, the following characteristic factors are calculated for each range library in the radar scanning radial direction: echo standard deviation (STDEV), horizontal reflectivity factor structure (TDBZ), reflectivity factor horizontal texture (Texture), reflectivity factor vertical gradient (VertGrad), echo library variation (SPIN), echo sign change along the radial direction (SIGN), radial velocity region average (MDVE), velocity variance (SDVE), spectral width region average (MDSW), and echo top height (EchoTop).

[0077] (1) Echo standard deviation ZSTDEV

[0078]

[0079] N: Represents the number of echo points in the template, X i : Echo intensity with subscript i The average value of the echo intensity within the template.

[0080] (2) Horizontal reflectivity factor structure TDBZ

[0081]

[0082] Nrays and Ngates: These represent the calculation ranges defined along the azimuth and radial directions, respectively. i.j This represents the value of the reflectivity factor.

[0083] (3) Reflectivity Factor Horizontal Texture

[0084]

[0085] (4) Inter-column variation of echo along the radial direction SPIN

[0086]

[0087]

[0088] Zthresh represents the threshold for changes in inter-column echo intensity, typically ranging from 2 to 5 dBZ.

[0089] (5) Intercolumn sign change of the echo along the radial direction (SIGN)

[0090]

[0091]

[0092] (6) Vertical gradient of reflectivity factor VertGrad

[0093]

[0094] Z up and Z low θ represents the echo values ​​of the corresponding echo point at the high elevation angle layer and the low elevation angle layer, respectively. up θ low These represent the elevation angle values ​​for high-rise and low-rise buildings, respectively.

[0095] (7) Median radial velocity MDVE: The radial velocity value after median filtering of the echo point.

[0096] (8) Velocity Variance SDVE

[0097]

[0098] X i.j This represents the radial velocity value of the echo points within the selected area. The median velocity within the selected area.

[0099] (9) Median velocity spectrum width MDSW: The velocity spectrum width value after median filtering of the echo point.

[0100] (10) EchoTop

[0101]

[0102] Z t Echo threshold (5dBZ), Z h and Z l H represents the echo intensity values ​​corresponding to the elevation angles of the upper and lower floors, respectively. h and H l These represent the echo heights of the echo point at the corresponding high and low elevation angles, respectively. The echo height H can be obtained using the altimeter formula.

[0103]

[0104] Where h0 represents the radar antenna height, R represents the slant range between the target and the radar, α represents the elevation angle, and Rm is the equivalent Earth radius.

[0105] Step S4: Determine the threshold values ​​of the discriminant characteristic parameters for different types of radar echoes.

[0106] Different types of radar echoes, including precipitation, clear sky, ground features, and super-refractive data, were collected. Based on statistical analysis of a large number of case studies, the discriminant factors for different types of radar echoes are as follows:

[0107] (1) Ground features and super-refractive echo: The horizontal reflectivity factor structure TDBZ and horizontal texture TEXTURE are large, the SPIN value is relatively large, the vertical gradient of reflectivity factor VertGrad is less than 0 and the absolute value is very large, the median radial velocity MDVE is close to 0, the velocity variance SDVE and the median velocity spectrum width MDSW are very small, and the EchoTop is between 3-4.5 km.

[0108] (2) Clear sky echo: The echo intensity is relatively small, generally less than 15dBZ; the echo standard deviation ZSTDEV is small, the SPIN value is small, and the EchoTop is generally less than 3km.

[0109] (3) Precipitation echo: TDBZ and TEXTURE are generally less than 30, SPIN value is small, vertical gradient is closer to 0, and EchoTop is greater than 5km.

[0110] Step S5: Filter the discriminant factors for different radial range circles of the radar.

[0111] The effective radius of radar radial velocity data is generally around 150 km, with a maximum detection radius of 230 km. Therefore, we divide the statistical interval into three parts: 0-50 km, 50-150 km, and greater than 150 km. Within 0-50 km, we initially selected R, ZSTDEV, TDBZ, TEXTURE, VertGrad, SPIN, MDVE, SDVE, and MDSW as discriminant factors. Within 50-150 km, we initially selected R, ZSTDEV, TDBZ, TEXTURE, VertGrad, SPIN, MDVE, SDVE, MDSW, and EchoTop as discriminant factors. Within >150 km, we initially selected R, ZSTDEV, TDBZ, TEXTURE, VertGrad, SPIN, and EchoTop as discriminant factors.

[0112] Step S6: Establish a radar echo classification and discrimination model

[0113] Statistical analysis was conducted on a large number of radar echo cases of different types in Step S4. Discriminant factors for each interval were screened and determined. Stepwise regression was used to determine the influencing factors of the discriminant factors in different intervals. The maximum likelihood method was used to estimate the parameters of each discriminant factor, and a Logistic regression model for radar echo classification was established.

[0114] The logistic regression model is mainly used to predict the probability of events influenced by multiple factors. It represents the occurrence of an event as Y = 0, and the non-occurrence of an event as Y = 1. E(Y) = p represents the probability of an event occurring, X1…X… m These are the predictor factors; then, the coefficients of the Logistic model are calculated based on the sample data.

[0115] ln(p / (1-p))=b0+b1X1+...b m X m +ε=b0+β'X+ε

[0116] The maximum likelihood method is typically used to estimate b0b1…b m That is, by

[0117]

[0118] Solve for X, where n is the number of observations; k is the number of events; and X is the number of occurrences. i It is the observed value of the predictor factor when the event occurs, which is X. j Observed values ​​of the forecast factor when the event does not occur.

[0119] Specific examples

[0120] This example uses two weather events: winter precipitation at location A on November 26, 2022, and superrefractive precipitation at location B on April 10, 2009.

[0121] like Figure 3 As shown in the winter precipitation echo map of location A on November 26, 2022, there are weak precipitation echoes in the northwest corner 100-200 kilometers from the radar station center and in the northeast direction from the radar center, interspersed with some clear sky echoes. Figure 4 This refers to a strong super-refraction phenomenon that occurred at location B on April 10, 2009. Located approximately 80 kilometers southeast of the radar station, strong super-refraction began to occur and continued, with the strongest echo reaching 76 dBZ. As shown in the image, the echo model can effectively distinguish between precipitation and non-precipitation echoes. In practical applications, failing to remove these non-precipitation echoes would affect the accurate identification of weather conditions.

[0122] The above description is a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

[0123] This patent is not limited to the above-described preferred embodiments. Anyone can derive other forms of radar echo classification and identification methods based on multi-factor parameter features under the guidance of this patent. All equivalent changes and modifications made within the scope of this patent application shall fall within the scope of this patent.

Claims

1. A radar echo classification and identification method based on multi-factor parameter features, characterized in that, Includes the following steps: Step S1: Acquire the base data detected by CINRAD / SA weather radar; Step S2: Remove isolated noise interference from radar base data; Step S3: Calculate the feature factors used to classify radar echoes: For the radar base data after quality control in step S2, calculate radar echo characteristic factors including at least the echo standard deviation ZSTDEV, horizontal reflectivity factor structure TDBZ, reflectivity factor horizontal texture, reflectivity factor vertical gradient VertGrad, echo inter-column variation SPIN, echo radial sign change SIGN, radial velocity region average MDVE, velocity variance SDVE, spectral width region average MDSW, and echo top height EchoTop. Step S4: Based on statistical analysis of individual radar echo cases of different types, determine the threshold values ​​of the discriminant factor characteristic parameters for different types of radar echoes; Step S5: Select the discrimination factors for different radial range circles of the radar. Specifically: in the 0-50 km range, R, ZSTDEV, TDBZ, Texture, VertGrad, SPIN, MDVE, SDVE, and MDSW are used as discrimination factors; in the 50-150 km range, R, ZSTDEV, TDBZ, Texture, VertGrad, SPIN, MDVE, SDVE, MDSW, and EchoTop are used as discrimination factors; in the >150 km range, R, ZSTDEV, TDBZ, Texture, VertGrad, SPIN, and EchoTop are used as discrimination factors. Step S6: Based on step S5, determine the influencing factors of the discrimination factors for each interval, and establish a radar echo classification and discrimination model.

2. The radar echo classification and recognition method based on multi-factor parameter features according to claim 1, characterized in that: In step S1, the CINRAD / SA weather radar base data is arranged from low to high elevation angle: 0.5-19.5 degrees, the elevation angle scan data of each layer is arranged according to radial scan lines 0-360 degrees, and each radial scan line is stored in polar coordinates according to a distance library of 1 km × 1 degree.

3. The radar echo classification and identification method based on multi-factor parameter features according to claim 1, characterized in that: In step S2, the removal of isolated noise from radar base data is achieved by taking a selected echo point as the center and comparing the percentage of invalid reflectivity factor values ​​P within a 5×5 window with a threshold to remove isolated noise interference.

4. The radar echo classification and identification method based on multi-factor parameter features according to claim 1, characterized in that: In step S3, the specific calculation methods for the characteristic factors used to classify radar echoes, including echo standard deviation ZSTDEV, horizontal reflectivity factor structure TDBZ, reflectivity factor horizontal texture, reflectivity factor vertical gradient VertGrad, echo inter-column variation SPIN, echo radial sign change SIGN, radial velocity region average MDVE, velocity variance SDVE, spectral width region average MDSW, and echo top height EchoTop, are as follows: (1) Echo standard deviation ZSTDEV N: Represents the number of echo points in the template. : Echo intensity with subscript i The average value of the echo intensity within the template; (2) Horizontal reflectivity factor structure TDBZ Nrays and Ngates: These represent the calculation ranges defined along the azimuth and radial directions, respectively. This represents the value of the reflectivity factor; (3) Reflectivity factor horizontal texture (4) Inter-column variation of echo along the radial direction SPIN Zthresh represents the threshold for changes in inter-column echo intensity; (5) Intercolumn sign change of the echo along the radial direction (SIGN) (6) Vertical gradient of reflectivity factor VertGrad and These represent the echo values ​​of the corresponding echo points at high elevation angle and low elevation angle, respectively. , These represent the elevation angle values ​​for the high-rise and low-rise buildings, respectively. (7) Median Radial Velocity MDVE: Radial velocity value after median filtering of echo points; (8) Velocity Variance SDVE This represents the radial velocity value of the echo points within the selected area. The median velocity within the selected area; (9) Median velocity spectrum width MDSW: The velocity spectrum width after median filtering of the echo point; (10) EchoTop Echo threshold, 5dBZ and These represent the echo intensity values ​​corresponding to the elevation angles of the higher and lower floors, respectively. and These represent the echo point heights at the corresponding high and low elevation angles, respectively; the echo height H is obtained using the altimeter formula. in, R represents the radar antenna height, R represents the slant distance between the target and the radar, α represents the elevation angle, and Rm is the equivalent Earth radius.

5. The radar echo classification and identification method based on multi-factor parameter features according to claim 1, characterized in that: In step S6, the influencing factors of the discriminant factors for each interval are determined and a Logistic regression model is established.

Images