Weather radar data quality control method based on dynamic clutter map and texture features

Abstract

The present invention discloses a weather radar data quality control method based on dynamic clutter maps and texture features, which relates to the field of meteorological radar technology. The method comprises the following steps: inputting radar volume scan data; calculating clutter frequency statistics; generating a mask matrix based on the clutter map and reflection threshold, preliminarily determining the clutter location based on the mask matrix, and preliminarily filtering out the clutter; defining a 9*9 window to slide and capture the scan elevation data, extracting the velocity characteristics and spectral width texture characteristics of the clutter area at each layer of the scan elevation, and calculating the velocity standard deviation and spectral width texture standard deviation; performing a secondary determination on the preliminarily filtered out clutter based on the velocity standard deviation and spectral width texture standard deviation, filtering out echoes that are secondary determined to be clutter, and retaining other echoes that are not secondary determined to be clutter; filling holes; and outputting the processed radar data. The present invention improves the accuracy and reliability of weather radar data, thereby providing higher-quality data support for meteorological monitoring and forecasting.

Description

Technical Field

[0001] The present invention relates to the technical field of weather radar, and more particularly to a weather radar data quality control method based on dynamic clutter maps and texture features. Background Art

[0002] With the continuous advancement of meteorological monitoring technology, weather radar has become a vital tool for monitoring and predicting weather changes. Weather radar detects meteorological targets such as precipitation and storms in the atmosphere by emitting electromagnetic waves and receiving reflected signals. However, in practical applications, weather radar is subject to interference from various factors, resulting in degraded data quality. These interference factors primarily include ground clutter, sea clutter, interference from insects and birds, and system noise. Ground clutter is a major issue in weather radar data quality control.

[0003] Ground clutter refers to the reflected signals generated by the radar beam during scanning, when it strikes fixed or slow-moving objects such as the ground, buildings, and mountains. These clutter signals mix with the meteorological target signals, affecting the accuracy of meteorological data and reducing the effectiveness and reliability of weather radar in precipitation estimation, storm detection, and weather forecasting.

[0004] Traditional weather radar data quality control methods mainly include fixed threshold methods, adaptive filtering methods, and clutter map methods. The fixed threshold method distinguishes between meteorological targets and clutter by setting a fixed reflectivity or velocity threshold, but this method has limitations when dealing with clutter under different environmental conditions. The adaptive filtering method dynamically adjusts the filter parameters based on the statistical characteristics of the signal, which can better adapt to different clutter environments, but its computational complexity is high and its real-time performance is poor. The clutter map method uses pre-established ground clutter maps to identify and remove clutter signals, but traditional clutter maps are usually generated based on historical data statistics. They are not flexible and accurate enough when dealing with ground changes or short-term clutter changes. In addition, traditional clutter map methods often filter out meteorological echoes incorrectly.

[0005] In summary, existing weather radar data quality control methods have certain limitations when dealing with complex and changing ground clutter and cannot meet the needs of high-precision meteorological monitoring. Therefore, it is particularly urgent to develop a new method that can effectively improve the quality of weather radar data. Summary of the Invention

[0006] The present invention aims to address the above-mentioned technical problems by providing a weather radar data quality control method based on dynamic clutter patterns and texture features. This method overcomes the shortcomings of existing technologies, improves the accuracy and reliability of weather radar data, and thus provides higher-quality data support for meteorological monitoring and forecasting.

[0007] In order to achieve the above-mentioned purpose, the present invention specifically adopts the following technical solutions:

[0008] The present invention provides a weather radar data quality control method based on dynamic clutter map and texture features, comprising the following steps:

[0009] S1. Radar volume scan data input (radar volume scan data refers to data obtained by weather radar volume scanning. This scanning method not only scans near the ground but also rises at multiple elevation angles to obtain vertical data).

[0010] S2. Clutter frequency statistics: Count the frequency of unfiltered reflectivity in the clutter map. If the statistical number is greater than the set value N, proceed to the next step.

[0011] S3. Use linear fitting to calculate the frequency change gradient, find the jump point in the change gradient, determine the frequency probability threshold, update the clutter map, and clear the statistical frequency;

[0012] S4. Generate a mask matrix based on the clutter map and the reflection threshold, preliminarily determine the clutter position based on the mask matrix, and preliminarily filter out the clutter;

[0013] S5. Define a 9*9 window to slide and grab the scanning elevation data, extract the velocity characteristics and spectral width texture characteristics of the clutter area of ​​each layer of scanning elevation, and calculate the velocity standard deviation and spectral width texture standard deviation;

[0014] S6. Performing a secondary determination on the clutter initially filtered out in step S4 based on the velocity standard deviation and the spectral width texture standard deviation. Echoes that meet the clutter map condition and the spectral width texture condition are secondary determined to be clutter. Echoes secondary determined to be clutter are filtered out, and other echoes not secondary determined to be clutter are retained.

[0015] S7, filtering out holes in the clutter map of the echoes that are secondarily determined to be clutter, and filling the holes;

[0016] S8. Output the processed radar data.

[0017] In one embodiment, in step S2, the radar volume scan data is the original reflectivity data before radar signal processing. The original reflectivity data retains clutter such as ground object echoes, water vapor echoes, and biological echoes to the greatest extent. The clutter frequency statistics are performed using the clutter map of the unfiltered reflectivity. The specific method of clutter frequency statistics is as follows:

[0018] S21. After the radar completes a volume scan, each layer of scan elevation angle data has n radials, and each radial has m range bins. First, calculate the maximum radial number MaxRd and the maximum range bin number MaxBin of each layer of scan elevation angle layer by layer.

[0019] S22. Define an initial matrix MaskInit of size MaxRd*MaxBin for each layer of scanning elevation angle, and initialize all values ​​of the matrix to 0. Traverse the unfiltered reflectivity data of each layer of scanning elevation angle radially and range-bin by range. If the current position is a valid value, add 1 to the value of the corresponding position of MaskInit. When the number of statistics is greater than N (N can be adjusted according to the radar scanning speed), obtain the clutter frequency map.

[0020] In one embodiment, in step S3, linear fitting is used to calculate the frequency change gradient and determine the probability threshold, as follows: the frequency graph counts the number of times a valid value appears at each location, but the valid value includes normal weather echoes and clutter that needs to be filtered out. Therefore, the result of the frequency graph is to count both weather echoes and clutter at the same time. In order to distinguish between the two, a probability graph is first calculated based on the frequency graph.

[0021] In one embodiment, a probability map is first calculated based on the frequency map. The specific method is as follows:

[0022] S31. When counting the frequency graph, maintain a count number Cnt, and add 1 to Cnt for each volume scan data counted; finally, divide each value in the initial matrix MaskInit by Cnt, and calculate its probability value to obtain the clutter probability matrix Maskprob, which is the clutter probability graph used later;

[0023] S32. After obtaining the clutter probability matrix Maskprob, all values ​​in the clutter probability matrix Maskprob are expanded into a one-dimensional array and sorted in ascending order to obtain Sortprop. Then, a window of length 500 is used to traverse Sortprop starting from 0. While traversing, a linear fit is used to fit the 500 data in the window into a straight line of the shape y=kx+b, and the slope k of the straight line is recorded.

[0024] S33. When the window completes the traversal of Sortprop, a curve of the change in the slope of the fitting line is obtained, and the window starting position MaxKLoc corresponding to the maximum slope is found. The probability value corresponding to MaxKLoc is found in Sortprop as the frequency probability threshold finalP.

[0025] Specifically, the Sortprop plotting results and the determination of the frequency probability threshold, the coordinates of the intersection of the red dotted line and the blue probability curve marked in the figure are (MaxKLoc, finalP).

[0026] In one embodiment, in step S4, the clutter position is preliminarily determined in the following manner:

[0027] Based on the clutter probability matrix Maskprob and the frequency probability threshold finalP, the echoes are initially filtered out: a mask matrix Judge is generated with the same size as the clutter probability matrix Maskprob. For each position, if the probability value of the clutter probability matrix Maskprob is greater than the clutter frequency probability threshold finalP, the mask matrix Judge is assigned a value of True at the corresponding position; otherwise, it is assigned a value of False. The echoes whose corresponding positions in the mask matrix Judge are False are filtered out.

[0028] Specifically, Figure 6 Schematic diagram of the mask matrix Judge; Figure 7 is the original reflectivity map; Figure 8 This is the reflectivity map after preliminary filtering.

[0029] In one embodiment, in step S6, when precipitation occurs, the reflectivity after preliminary filtering may be significantly misfiltered. In particular, when precipitation echoes pass close to the radar, they may be misidentified as clutter. This is because the radar often detects clear-air echoes such as water vapor and aerosols at close range. These echoes significantly interfere with the statistics of the clutter map, resulting in a large number of misfilters during preliminary filtering. To address this misfiltering, it is necessary to retain normal meteorological echoes. The specific method is as follows:

[0030] Based on the velocity and spectral width data of the same elevation layer as the reflectivity, a 9*9 window is used to traverse the velocity and spectral width data over the entire scanning elevation angle, and the standard deviation of the data in each window is calculated to extract the corresponding texture features.

[0031] Since the texture features of the speed and spectral width of the clutter are more obvious than those of the weather echo, which is reflected in the value of the standard deviation as the standard deviation of the speed and spectral width of the clutter is larger, two threshold values ​​can be obtained based on this: the speed texture threshold value V_limit and the spectral width texture threshold value W_limit. Among the echoes filtered out in step S4, the echoes that meet the speed texture feature less than V_limit and the spectral width texture feature less than W_limit will be retained.

[0032] Specifically, meteorological echoes are generated by the reflection of radar waves from particles in the atmosphere, such as cloud droplets, raindrops, and ice crystals. These particles are affected by airflow, precipitation type and intensity, and phase changes, resulting in complex and irregular spatial distribution and motion. For example, in cumulonimbus clouds, strong updrafts and downdrafts cause significant variations in particle distribution and motion, resulting in a wide range of velocity variations and large spectral width variations, leading to significant texture characteristics. In contrast, clutter, such as ground objects, sea clutter, and insects, has a narrower range of velocity and spectral width variations and smaller texture characteristics.

[0033] Different radar systems and emission wavelengths have different threshold values, depending on the texture calculation results. The threshold values ​​used in this paper are: V_limit = 2.5, W_limit = 1.8. Among the echoes filtered out in step S4, those with velocity texture characteristics less than V_limit and spectral width texture characteristics less than W_limit are retained. By calculating texture characteristics, most weather echoes can be well preserved.

[0034] In one embodiment, in step S7, the specific method for filling the filtered hole is: first determine the boundary position of the hole, then select three adjacent radials around the hole boundary in azimuth, select reflectivity data of three distance libraries around the hole boundary in distance, calculate the median of these data, and use the median as the filling point; then, repeat the operation of selecting three adjacent radials in azimuth, selecting reflectivity data of three distance libraries around the hole boundary in distance and calculating the median filling at the new boundary position inside the hole, until the filling of the entire hole center position is completed.

[0035] Specifically, after the initial filtering in step S4 and the echo retention in step S6, a very small amount of weather echoes may still be mistakenly filtered out, forming some holes. To fill these holes, the principle of neighborhood approximation is used to fill the holes from the outside to the inside.

[0036] The beneficial effects of the present invention are as follows:

[0037] The present invention is rationally designed and can improve the accuracy and reliability of weather radar data, thereby providing higher-quality data support for meteorological monitoring and forecasting. The specific effects are as follows:

[0038] Limitations of the fixed threshold method have been overcome: The traditional fixed threshold method is difficult to adapt to clutter in different environments due to the constant reflectivity or velocity threshold, which often results in some clutter remaining or weather echoes being mistakenly filtered out. The present invention determines the probability threshold by counting the clutter frequency and combining linear fitting to calculate the frequency change gradient, dynamically distinguishes meteorological echoes from clutter, adapts to a variety of environmental conditions, accurately filters out clutter, and avoids false filtering. The fixed threshold method is difficult to take into account the identification of meteorological echoes under different precipitation types and intensities. The present invention uses velocity texture features and spectral width texture features for secondary judgment, and adaptively retains the true meteorological echo according to the differences in echo characteristics in different precipitation scenes, effectively solving the shortcomings of the fixed threshold method under complex precipitation conditions.

[0039] Optimizing Adaptive Filtering: While traditional adaptive filtering methods can dynamically adjust filter parameters to adapt to varying clutter environments, they suffer from high computational complexity and poor real-time performance, making them difficult to apply in real-world applications. While maintaining effective filtering, this new method constructs a dynamic clutter map through clutter frequency statistics and probability map calculations. This significantly reduces computational effort, enables rapid radar data processing, meets real-time meteorological monitoring requirements, and improves data processing efficiency.

[0040] Improvements on the traditional clutter map method: The traditional clutter map relies on historical data statistics to generate, which is not flexible and accurate enough when facing changes in ground objects or short-term clutter changes, and is prone to falsely filtering out meteorological echoes. The clutter map of the present invention is dynamically updated. After the statistical frequency is greater than the set value, the new threshold is determined based on the frequency change gradient jump point and the clutter map is updated, so that the clutter map reflects the current clutter situation in real time, enhances the adaptability to clutter changes, and reduces false filtering. After filtering out clutter, the traditional clutter map method is prone to causing holes in the data, affecting data integrity. The unique hole-filling method of the present invention fills the holes with the median of the reflectivity data around the hole boundary, gradually advancing from the outside to the inside, effectively ensuring the integrity of radar data, avoiding omissions of meteorological information due to missing data, and improving the reliability of meteorological monitoring and forecasting.

[0041] Improved Comprehensive Benefits: In terms of precipitation estimation and storm detection, this invention effectively improves the accuracy of precipitation estimation and storm detection by precisely filtering out clutter and retaining meteorological echoes. This provides more reliable data to meteorological departments, enhances the ability to warn of severe weather events, and significantly improves the quality and effectiveness of meteorological services compared to traditional methods. This invention is applicable to a variety of radar systems and has strong versatility. It can be widely used to control the quality of data from different types of weather radars, providing high-quality data resources for meteorological research, promoting the development of meteorological disciplines, and helping to address global issues such as climate change. It has broad application prospects in the meteorological field. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 It is a flow chart of the weather radar data quality control method based on dynamic clutter map and texture features of the present invention;

[0044] Figure 2 is the clutter frequency.

[0045] Figure 3 is the clutter probability map.

[0046] Figure 4 This is the fitting slope change curve.

[0047] Figure 5 It is a probability curve graph.

[0048] Figure 6 Schematic diagram of the mask matrix Judge.

[0049] Figure 7 is the original reflectivity map.

[0050] Figure 8 This is the reflectivity map after preliminary filtering.

[0051] Figure 9 The original speed graph.

[0052] Figure 10 For the velocity texture map.

[0053] Figure 11 is the original spectrum width.

[0054] Figure 12 is the spectrum width texture map.

[0055] Figure 13 It is the reflectivity map after preserving the texture features.

[0056] Figure 14 Schematic diagram of hole filling.

[0057] Figure 15 This is the reflectivity map after the holes are filled. DETAILED DESCRIPTION

[0058] To make the technical problems, technical solutions, and technical effects of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0059] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0060] Example 1

[0061] like Figure 1 As shown, this embodiment provides a weather radar data quality control method based on dynamic clutter map and texture features, including the following steps:

[0062] S1. Radar volume scan data input (radar volume scan data refers to data obtained by weather radar volume scanning. This scanning method not only scans near the ground but also rises at multiple elevation angles to obtain vertical data).

[0063] S2. Clutter frequency statistics: Radar volume scan data is the original reflectivity data before radar signal processing. The original reflectivity data retains clutter such as ground object echo, water vapor echo, and biological echo to the greatest extent. Clutter frequency statistics are performed using the clutter map of the unfiltered reflectivity. The specific method for clutter frequency statistics is as follows:

[0064] S21. After the radar completes a volume scan, each layer of scan elevation angle data has n radials, and each radial has m range bins. First, calculate the maximum radial number MaxRd and the maximum range bin number MaxBin of each layer of scan elevation angle layer by layer.

[0065] S22. Define an initial matrix MaskInit of size MaxRd*MaxBin for each layer of scanning elevation angle, and initialize all values ​​of the matrix to 0. Traverse the unfiltered reflectivity data of each layer of scanning elevation angle radially and range-by-range bin one by one. If the current position is a valid value, then add 1 to the value of the corresponding position of MaskInit. When the number of statistics is greater than N (N can be adjusted according to the radar scanning speed), obtain the clutter frequency map, as shown in the following example: Figure 2 shown.

[0066] S3. Use linear fitting to calculate the frequency change gradient and determine the probability threshold. The details are as follows: The frequency graph counts the number of valid values ​​at each location, but the valid values ​​include normal weather echoes and clutter that needs to be filtered out. Therefore, the frequency graph results in the simultaneous counting of weather echoes and clutter. To distinguish between the two, first calculate the probability graph based on the frequency graph. The specific method is as follows:

[0067] S31. When counting the frequency graph, maintain a count number Cnt. Each time a volume scan data is counted, Cnt is increased by 1. Finally, each value in the initial matrix MaskInit is divided by Cnt, and its probability value is calculated to obtain the clutter probability matrix Maskprob, which is the clutter probability graph used later. The result is plotted as follows: Figure 3 ;

[0068] S32. After obtaining the clutter probability matrix Maskprob, expand all values ​​in the clutter probability matrix Maskprob into a one-dimensional array and sort them in ascending order to obtain Sortprop. Then, use a window of length 500 to traverse Sortprop starting from 0. While traversing, use linear fitting to fit the 500 data in the window into a straight line of shape y=kx+b, and record the slope k of the straight line ( Figure 4is the fitting slope change curve);

[0069] S33. When the window completes the traversal of Sortprop, a curve of the change in the slope of the fitting line is obtained, and the window starting position MaxKLoc corresponding to the maximum slope is found. The probability value corresponding to MaxKLoc is found in Sortprop as the frequency probability threshold finalP.

[0070] Specifically, Figure 5 This is the probability curve graph, the Sortprop drawing result and the determination of the frequency probability threshold. The coordinates of the intersection of the red dotted line and the blue probability curve marked in the figure are (MaxKLoc, finalP).

[0071] S4. Generate a mask matrix based on the clutter map and the reflection threshold, preliminarily determine the clutter location based on the mask matrix, and preliminarily filter out the clutter, as follows:

[0072] Based on the clutter probability matrix Maskprob and the frequency probability threshold finalP, the reflectivity is initially filtered out: a mask matrix Judge is generated with the same size as the clutter probability matrix Maskprob. For each position, if the probability value of the clutter probability matrix Maskprob is greater than the clutter frequency probability threshold finalP, the mask matrix Judge is assigned a value of True at the corresponding position; otherwise, it is assigned a value of False. The echoes whose reflectivity corresponding positions are False in the mask matrix Judge are filtered out.

[0073] Specifically, Figure 6 Schematic diagram of the mask matrix Judge; Figure 7 is the original reflectivity map; Figure 8 This is the reflectivity map after preliminary filtering.

[0074] S5. Define a 9*9 window to slide and grab the scanning elevation data, extract the velocity characteristics and spectral width texture characteristics of the clutter area of ​​each layer of scanning elevation, and calculate the velocity standard deviation and spectral width texture standard deviation;

[0075] S6. When encountering precipitation, the reflectivity after preliminary filtering will have obvious false positives. In particular, when precipitation echoes pass close to the radar, they will be mistakenly identified as clutter. This is because the radar often detects clear-air echoes such as water vapor and aerosols at close range. These echoes greatly interfere with the statistics of the clutter map, resulting in a large number of false negatives during preliminary filtering. To address the above false negatives, the clutter initially filtered out in step S4 is re-determined based on the velocity standard deviation and the spectral width texture standard deviation. Echoes that meet the clutter map conditions and spectral width texture conditions are re-determined as clutter. The echoes re-determined as clutter are filtered out, and the other echoes not re-determined as clutter are retained, as follows:

[0076] Based on the velocity and spectral width data of the same elevation layer as the reflectivity, a 9*9 window is used to traverse the velocity and spectral width data over the entire scanning elevation angle, and the standard deviation of the data in each window is calculated to extract the corresponding texture features.

[0077] Since the texture features of the speed and spectral width of the clutter are more obvious than those of the weather echo, which is reflected in the value of the standard deviation as the standard deviation of the speed and spectral width of the clutter is larger, two threshold values ​​can be obtained based on this: the speed texture threshold value V_limit and the spectral width texture threshold value W_limit. Among the echoes filtered out in step S4, the echoes that meet the speed texture feature less than V_limit and the spectral width texture feature less than W_limit will be retained.

[0078] The principle is as follows: Meteorological echoes are generated by the reflection of radar waves by particles such as cloud droplets, raindrops, and ice crystals in the atmosphere. These particles are affected by airflow, precipitation type and intensity, phase changes, etc., and their spatial distribution and movement are complex and irregular. For example, in cumulonimbus clouds, strong updrafts and downdrafts cause huge differences in particle distribution and movement, a wide range of velocity variations, and large spectral width variations, resulting in significant texture features. In contrast, clutter such as ground objects, sea clutter, and insects has a small range of velocity and spectral width variations, and smaller texture features. Figure 9 is the original velocity graph; Figure 10 is the velocity texture map; Figure 11 is the original spectrum width graph, Figure 12 is the spectrum width texture map.

[0079] Different radar systems have different emission wavelengths, and the threshold values ​​will vary depending on the texture calculation results. Here are the threshold values ​​​​given in this article: V_limit = 2.5, W_limit = 1.8. Among the echoes filtered out in step S4, the echoes with velocity texture characteristics less than V_limit and spectral width texture characteristics less than W_limit will be retained. Figure 13 is the reflectivity map after texture feature preservation, from Figure 13 It can be seen that by calculating the texture features, most weather echoes can be well preserved.

[0080] S7. There are holes in the clutter map after filtering out the echoes that are secondarily determined to be clutter, and the holes are filled. The specific method is: first determine the boundary position of the hole, then select three adjacent radials around the hole boundary in azimuth, select reflectivity data of three distance bins around the hole boundary in distance, calculate the median of these data, and use the median as the filling point; then, repeat the operation of selecting three adjacent radials in azimuth and three distance bins in distance and calculating the median to fill the hole at the new boundary position inside the hole, until the center position of the entire hole is filled.

[0081] Although after the initial filtering in step S4 and the echo retention in step S6, a very small amount of weather echoes will still be mistakenly filtered out, forming some holes. To fill these holes, the principle of neighborhood approximation is used to fill the holes from the outside to the inside. Figure 14 Schematic diagram of hole filling. Figure 15 This is the reflectivity map after the holes are filled.

[0082] S8. Output the processed radar data.

Claims

1. A weather radar data quality control method based on dynamic clutter maps and texture features, characterized in that: The steps include: S1, radar volume scan data input; S2. Clutter frequency statistics: Count the frequency of unfiltered reflectivity in the clutter map. If the statistical number is greater than the set value N, proceed to the next step. S3. Use linear fitting to calculate the frequency change gradient, find the jump point in the change gradient, determine the frequency probability threshold, update the clutter map, and clear the statistical frequency; S4. Generate a mask matrix based on the clutter map and the reflection threshold, preliminarily determine the clutter position based on the mask matrix, and preliminarily filter out the clutter; S5. Define a 9*9 window to slide and grab the scanning elevation data, extract the velocity characteristics and spectral width texture characteristics of the clutter area of ​​each layer of scanning elevation, and calculate the velocity standard deviation and spectral width texture standard deviation; S6. Performing a secondary determination on the clutter initially filtered out in step S4 based on the velocity standard deviation and the spectral width texture standard deviation. Echoes that meet the clutter map condition and the spectral width texture condition are secondary determined to be clutter. Echoes secondary determined to be clutter are filtered out, and other echoes not secondary determined to be clutter are retained. S7, filtering out holes in the clutter map of the echoes that are secondarily determined to be clutter, and filling the holes; S8. Output the processed radar data.

2. The weather radar data quality control method based on dynamic clutter map and texture features according to claim 1 is characterized in that: In step S2, the specific method of clutter frequency statistics is as follows: S21. After the radar completes a volume scan, each layer of scan elevation angle data has n radials, and each radial has m range bins. First, calculate the maximum radial number MaxRd and the maximum range bin number MaxBin of each layer of scan elevation angle layer by layer. S22. Define an initial matrix MaskInit of size MaxRd*MaxBin for each layer of scanning elevation angle, and initialize all values ​​of the matrix to 0. Traverse the unfiltered reflectivity data of each layer of scanning elevation angle radially and range-by-range bin one by one. If the current position is a valid value, add 1 to the value of the corresponding position of MaskInit. When the number of statistics is greater than N, obtain the clutter frequency map.

3. The weather radar data quality control method based on dynamic clutter map and texture features according to claim 2 is characterized in that: In step S3, linear fitting is used to calculate the frequency change gradient and determine the probability threshold, as follows: S31. When counting the frequency graph, maintain a count number Cnt, and add 1 to Cnt for each volume scan data counted; finally, divide each value in the initial matrix MaskInit by Cnt, and calculate its probability value to obtain the clutter probability matrix Maskprob; S32. After obtaining the clutter probability matrix Maskprob, all values ​​in the clutter probability matrix Maskprob are expanded into a one-dimensional array and sorted in ascending order to obtain Sortprop. Then, a window of length 500 is used to traverse Sortprop starting from 0. While traversing, a linear fit is used to fit the 500 data in the window into a straight line of the shape y=kx+b, and the slope k of the straight line is recorded. S33. When the window completes the traversal of Sortprop, a curve of the change in the slope of the fitting line is obtained, and the window starting position MaxKLoc corresponding to the maximum slope is found. The probability value corresponding to MaxKLoc is found in Sortprop as the frequency probability threshold finalP.

4. The weather radar data quality control method based on dynamic clutter map and texture features according to claim 3 is characterized in that: In step S4, the clutter position is initially determined in the following manner: Based on the clutter probability matrix Maskprob and the frequency probability threshold finalP, the echoes are initially filtered out: a mask matrix Judge is generated with the same size as the clutter probability matrix Maskprob. For each position, if the probability value of the clutter probability matrix Maskprob is greater than the clutter frequency probability threshold finalP, the mask matrix Judge is assigned a value of True at the corresponding position; otherwise, it is assigned a value of False. The echoes whose corresponding positions in the mask matrix Judge are False are filtered out.

5. The weather radar data quality control method based on dynamic clutter map and texture features according to claim 4 is characterized in that: When encountering precipitation, the echoes after initial filtering may be significantly misfiltered. When precipitation echoes pass close to the radar, they are mistakenly identified as clutter. To resolve this misfiltering issue, it is necessary to retain normal weather echoes. The specific method is as follows: Based on the velocity and spectral width data of the same elevation layer as the reflectivity, a 9*9 window is used to traverse the velocity and spectral width data over the entire scanning elevation angle, and the standard deviation of the data in each window is calculated to extract the corresponding texture features. Since the texture features of the speed and spectral width of the clutter are more obvious than those of the weather echo, which is reflected in the value of the standard deviation as the standard deviation of the speed and spectral width of the clutter is larger, two threshold values ​​can be obtained based on this: the speed texture threshold value V_limit and the spectral width texture threshold value W_limit. Among the echoes filtered out in step S4, the echoes that meet the speed texture feature less than V_limit and the spectral width texture feature less than W_limit will be retained.

6. The weather radar data quality control method based on dynamic clutter map and texture features according to claim 5 is characterized in that: In step S7, the specific method for filling the filtered holes is as follows: first, determine the boundary position of the hole, then select three adjacent radials around the hole boundary in azimuth, select the reflectivity data of three distance libraries around the hole boundary in distance, calculate the median of these data, and use the median as the filling point; then, repeat the operation of selecting the reflectivity data of the three adjacent radials in azimuth and the three distance libraries around in distance and calculating the median filling at the new boundary position inside the hole until the center position of the entire hole is filled.

Images