Meteorological radar signal processor and meteorological radar signal processing program
By applying a window function and adjusting Doppler velocity width based on statistical properties, the technology enhances ground surface clutter removal and rainfall signal recovery in weather radar, addressing resolution and computational load issues.
Patent Information
- Application Number
- JP2023082291
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-18
- Publication Date
- 2024-11-28
- Estimated Expiration
- 2043-05-18
AI Technical Summary
Existing weather radar signal processing technologies face challenges in effectively removing ground surface clutter components while maintaining azimuthal beam resolution and avoiding excessive or incomplete removal of rainfall signal components, leading to increased computational load.
Applying a window function to time-series weather radar data and adjusting the Doppler velocity width based on statistical properties and observation points to enhance ground surface clutter removal and rainfall signal recovery, while reducing computational load.
Achieves effective removal of ground surface clutter, improves azimuthal beam resolution, and reduces the risk of rainfall signal component loss or over-recovery, while minimizing processing burden.
Smart Images

Figure 2024165796000001_ABST
Abstract
Description
[Technical field]
[0001] The present disclosure relates to techniques for removing ground clutter components from weather radar signals. [Background technology]
[0002] Known techniques for removing ground clutter components from weather radar signals include a technique for performing frequency domain filtering (MTI: Moving Target Indicator) and a technique for performing time domain filtering (GMAP-TD: Gaussian Model Adaptive Processing in Time Domain). GMAP-TD is disclosed in Non-Patent Document 1.
[0003] In MTI processing, the time series data signal is Fourier transformed into a Doppler velocity spectrum, ground clutter components near 0 Doppler velocity are removed, and rain signal components with finite Doppler velocity are extracted. However, if infinite length data approximation is applied to perform Fourier transform, the Doppler velocity spectrum deteriorates. Even if the main lobe of the ground clutter components is removed, if the side lobes of the ground clutter components remain, it becomes difficult to extract the rain signal components. Furthermore, if a window function is multiplied and MTI processing is performed, the side lobes of the ground clutter components are suppressed, but the rain signal components are also suppressed, so the gain of MTI processing decreases.
[0004] In the GMAP-TD processing of Non-Patent Document 1, the autocorrelation matrix of the time series data signal is multiplied by a filtering matrix to remove ground surface clutter components near the Doppler velocity of 0, and extract rain signal components with a finite Doppler velocity. Here, the Doppler velocity spectrum of the ground surface clutter components is set to a Gaussian that spreads from the main lobe to the side lobe, and the filtering matrix is calculated. This solves the problems of the MTI processing (deterioration of the Doppler velocity spectrum, difficulty in extracting rain signal components, and reduction in the gain of the MTI processing). However, if the ground surface clutter components near the Doppler velocity of 0 are removed, some rain signal components near the Doppler velocity of 0 are removed. Therefore, the Doppler velocity spectrum of some rain signal components with a finite Doppler velocity is fitted with a Gaussian with the average Doppler velocity as the distribution center, and some rain signal components near the Doppler velocity of 0 are restored. [Prior art documents] [Non-patent literature]
[0005] [Non-Patent Document 1] Cuong M. Nguyen and V. Chandrasekar, “Gaussian Model Adaptive Processing in Time Domain (GMAP-TD) for Weather Radars”, Journal of Atmospheric and Oceanic Technology, American Meteorological Society, November 2013, Volume 30, pp. 2571~2584. [Non-Patent Document 2] Sebastian Torres and Christopher Curtis, “Design considerations for improved tornado detection using super-resolution data on the NEXRAD network”, Preprints, Third European Conference on Radar Meteorology and Hydrology(ERAD), Barcelona, Spain, Copernicus.2006. Summary of the Invention [Problem to be solved by the invention]
[0006] In the beam resolution improvement process of Non-Patent Document 2, a window function is applied to time-series data of a weather radar signal in a weather radar device equipped with an antenna that rotates in the azimuth direction.
[0007] However, if the beam resolution improvement processing of Non-Patent Document 2 is applied in addition to the GMAP-TD processing of Non-Patent Document 1, the ground surface clutter components cannot be appropriately removed, and it is not possible to achieve both the effect of removing the ground surface clutter components and the effect of improving the azimuth beam resolution.
[0008] In the GMAP-TD processing of Non-Patent Document 1, when calculating a filtering matrix, the Doppler velocity width of the ground clutter components is set to a fixed width.
[0009] Therefore, when removing ground clutter components whose Doppler velocity width is originally small, if the fixed Doppler velocity width is too wide, there is a possibility that components other than ground clutter components will be removed in excess, and when removing tree clutter components whose Doppler velocity width is originally large, if the fixed Doppler velocity width is too narrow, there is a possibility that the tree clutter components will not be removed completely.
[0010] Furthermore, when extracting a rain signal component whose Doppler velocity is the mean near the Doppler velocity 0, if the Doppler velocity width of the rain signal component is narrow and the fixed Doppler velocity width is too wide, the rain signal component may be removed in excess, and the rain signal component may become unrecoverable. And when extracting a rain signal component whose Doppler velocity is the mean near the Doppler velocity 0, if the Doppler velocity width of the ground clutter component is wide and the fixed Doppler velocity width is too narrow, the remaining ground clutter component may be extracted as a spurious rain signal component, and the rain signal component may be over-recovered.
[0011] In the GMAP-TD process of Non-Patent Document 1, a filtering matrix is calculated during meteorological observation. Here, the calculation of the filtering matrix includes the calculation of an inverse matrix, as described later in Equation 14. Therefore, the calculation processing load increases during meteorological observation.
[0012] Therefore, in order to solve the above problems, the present disclosure aims to, when performing time-domain filtering processing on weather radar signals, (1) achieve both the effect of removing ground surface clutter components and the effect of improving azimuth beam resolution, (2) reduce the possibility that ground surface clutter is removed in excess or incompletely, and reduce the possibility that precipitation signal components become unrecoverable or over-recovered, and (3) reduce the computational processing burden during meteorological observation. [Means for solving the problem]
[0013] In order to solve the above problem, a window function is applied to the time series data of the weather radar signal, and the same window function is also applied to the time domain filtering process.
[0014] Specifically, the present disclosure relates to a weather radar signal processing device comprising: a weather radar signal acquisition unit that acquires time series data of a weather radar signal to which a window function has been applied from a weather radar device equipped with an antenna that rotates in an azimuth direction; a ground clutter removal unit that performs a time domain filtering process to which the window function has been applied to the time series data of the weather radar signal, when removing ground clutter components from the time series data of the weather radar signal to which the window function has been applied and extracting rainfall signal components; and a rainfall signal recovery unit that recovers the rainfall signal components removed together with the ground clutter components by the ground clutter removal unit, by calculating, based on statistical properties, a reception power intensity distribution with respect to the Doppler velocity of the rainfall signal components extracted by the ground clutter removal unit.
[0015] The present disclosure also relates to a weather radar signal processing program for causing a computer to sequentially execute a weather radar signal acquisition step of acquiring time series data of a weather radar signal to which a window function has been applied from a weather radar device equipped with an antenna that rotates in an azimuth direction; a ground surface clutter removal step of performing a time domain filtering process to which the window function has been applied to the time series data of the weather radar signal in order to remove ground surface clutter components from the time series data of the weather radar signal to which the window function has been applied and extract a rainfall signal component; and a rainfall signal recovery step of recovering the rainfall signal components removed together with the ground surface clutter components in the ground surface clutter removal step by calculating, based on statistical properties, a distribution of received power intensity with respect to the Doppler velocity of the rainfall signal components extracted in the ground surface clutter removal step.
[0016] If a window function is applied to the time series data of the weather radar signal, the statistical properties of the time series data of the weather radar signal, such as Gaussian, change. Therefore, unless the same window function is applied to the filtering process in the time domain as in Non-Patent Documents 1 and 2, the ground clutter components cannot be properly removed, and it is not possible to achieve both the effect of removing the ground clutter components and the effect of improving the azimuth beam resolution. However, if the same window function is applied to the filtering process in the time domain as in the present disclosure, the ground clutter components can be properly removed, and it is possible to achieve both the effect of removing the ground clutter components and the effect of improving the azimuth beam resolution.
[0017] The present disclosure also relates to a weather radar signal processing device, wherein the ground clutter removal unit sets a Doppler velocity width of the ground clutter components to be set in a time domain filtering process to which the window function is applied to the time series data of the weather radar signal, to a variable width depending on a meteorological observation point.
[0018] According to this configuration, when removing ground clutter components whose Doppler velocity width is originally small, the variable Doppler velocity width is appropriately narrowed, thereby reducing the possibility that components other than ground clutter components are removed in excess, and when removing tree clutter components whose Doppler velocity width is originally large, the variable Doppler velocity width is appropriately widened, thereby reducing the possibility that tree clutter components are not removed completely.
[0019] Furthermore, when extracting a rain signal component whose Doppler velocity average is near the Doppler velocity 0, even if the Doppler velocity width of the rain signal component is narrow, the variable Doppler velocity width can be appropriately narrowed to reduce the possibility that the rain signal component is removed excessively, and the possibility that the rain signal component becomes unrecoverable can be reduced. And, when extracting a rain signal component whose Doppler velocity average is near the Doppler velocity 0, even if the Doppler velocity width of the ground clutter component is wide, the variable Doppler velocity width can be appropriately widened to reduce the possibility that the remaining ground clutter component is extracted as a spurious rain signal component, and the possibility that the rain signal component is over-recovered can be reduced.
[0020] The present disclosure also relates to a weather radar signal processing device, wherein the ground clutter removal unit sets a Doppler velocity width of the ground clutter component to be set to a fixed width regardless of the meteorological observation point as an initial processing step in the filtering process of the time domain to which the window function applied to the time series data of the weather radar signal is applied, and the weather radar signal processing device further includes a discontinuity point extraction unit that extracts discontinuous observation points where at least one of the received power, Doppler velocity average, and Doppler velocity width of the rain signal component recovered by the rain signal recovery unit has discontinuity in a meteorological observation space, and the ground clutter removal unit changes the Doppler velocity width of the ground clutter component to be set to the variable width different from the fixed width at the discontinuous observation point extracted by the discontinuity point extraction unit as a post processing step in the filtering process of the time domain to which the window function applied to the time series data of the weather radar signal is applied.
[0021] According to this configuration, a meteorological observation point where the parameters of a rainfall signal component that is actually continuous erroneously have discontinuity (discontinuous observation point) is regarded as a meteorological observation point where the rainfall signal component is unrecoverable or over-recovered. Therefore, it is possible to distinguish between meteorological observation points where the rainfall signal component is unrecoverable and meteorological observation points where the rainfall signal component is actually small. Also, it is possible to distinguish between meteorological observation points where the rainfall signal component is over-recovered and meteorological observation points where the rainfall signal component is actually large.
[0022] Furthermore, in most of the meteorological observation points (continuous observation points) where the rainfall signal components are recoverable, the variable Doppler velocity width is set to only a fixed width. Then, in a portion of the meteorological observation points (discontinuous observation points) where the rainfall signal components are unrecoverable or over-recovered, the variable Doppler velocity width is changed to a variable width. This reduces the processing load for setting the variable Doppler velocity width.
[0023] The present disclosure also relates to a weather radar signal processing device, characterized in that the discontinuity extraction unit extracts, as the discontinuous observation points, meteorological observation points where the Doppler velocity average of the rainfall signal component recovered by the rainfall signal recovery unit has an inverted sign compared to surrounding meteorological observation points and where the received power of the rainfall signal component recovered by the rainfall signal recovery unit or the power difference before and after recovery is smaller than a predetermined difference or more compared to surrounding meteorological observation points, and the ground surface clutter removal unit changes, as a post-processing step, the Doppler velocity width of the ground surface clutter component to be set in a time domain filtering process to which the window function applied to the time series data of the weather radar signal is applied, to the variable width narrower than the fixed width at the discontinuous observation points extracted by the discontinuity extraction unit.
[0024] According to this configuration, when extracting a rain signal component whose Doppler velocity average is near 0 Doppler velocity, even when the Doppler velocity width of the rain signal component is narrow, by appropriately narrowing the variable Doppler velocity width, it is possible to reduce the possibility that the rain signal component is excessively removed and to reduce the possibility that the rain signal component becomes unrecoverable.
[0025] The present disclosure also provides a weather radar signal processing device, characterized in that the discontinuity extraction unit extracts, as the discontinuous observation points, meteorological observation points where the Doppler velocity average of the rainfall signal component recovered by the rainfall signal recovery unit has an inverted sign compared to surrounding meteorological observation points and where the received power of the rainfall signal component recovered by the rainfall signal recovery unit or the power difference before and after recovery is greater than or equal to a predetermined difference compared to surrounding meteorological observation points, and the ground surface clutter removal unit changes, as a post-processing step, the Doppler velocity width of the ground surface clutter component to be set in a time domain filtering process to which the window function applied to the time series data of the weather radar signal is applied, to the variable width which is wider than the fixed width at the discontinuous observation points extracted by the discontinuity extraction unit.
[0026] According to this configuration, when extracting rain signal components whose Doppler velocity average is near 0 Doppler velocity, even when the Doppler velocity width of the ground surface clutter components is wide, the variable Doppler velocity width is appropriately widened to reduce the possibility that the remaining ground surface clutter components are extracted as spurious rain signal components and reduce the possibility that the rain signal components are over-recovered.
[0027] The present disclosure also relates to a weather radar signal processing device, wherein the ground surface clutter removal unit stores, prior to meteorological observation, various time domain filtering characteristics to which the window function is applied to the time-series data of the weather radar signal in accordance with various reception intensities and Doppler velocity widths of ground surface clutter components.
[0028] According to this configuration, since the filtering matrix is stored before weather observation, the computational burden during weather observation can be reduced even when the calculation of the filtering characteristics includes the calculation of an inverse matrix, as described later in Equation 14.
[0029] The above disclosed inventions can be combined as much as possible. Effect of the Invention
[0030] In this way, when performing time-domain filtering on weather radar signals, the present disclosure can (1) achieve both the effect of removing ground surface clutter components and the effect of improving azimuth beam resolution, (2) reduce the possibility that ground surface clutter will be removed in excess or incompletely, and reduce the possibility that precipitation signal components will be unable to be recovered or will be over-recovered, and (3) reduce the computational burden during meteorological observation. [Brief description of the drawings]
[0031] [Figure 1] FIG. 1 is a diagram illustrating a problem with beam resolution enhancement processing in the prior art. [Diagram 2] FIG. 1 is a diagram showing an overview of a beam resolution enhancement process according to the prior art. [Diagram 3] FIG. 1 is a diagram illustrating a configuration of a weather radar signal processing device according to the present disclosure. [Figure 4] FIG. 2 is a diagram showing a procedure of ground surface clutter removal processing according to the present disclosure. [Diagram 5] FIG. 2 is a diagram showing an overview of the ground surface clutter removal process of the present disclosure. [Figure 6] 1A and 1B are diagrams illustrating problems with the filtering characteristic calculation process of the present disclosure. [Figure 7] FIG. 11 is a diagram showing a procedure of a filtering characteristic storage process according to the present disclosure. [Figure 8] 11A and 11B are diagrams illustrating the effect of the filtering characteristic calculation process of the present disclosure. [Figure 9] 11A and 11B are diagrams illustrating the effect of the filtering characteristic calculation process of the present disclosure. [Figure 10] FIG. 1 is a diagram showing an overview of a ground surface clutter removal range setting process according to the present disclosure. [Figure 11] FIG. 4 is a diagram showing a procedure of a ground surface clutter removal range setting process according to the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0032] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present disclosure will be described with reference to the accompanying drawings. The embodiments described below are examples of the present disclosure, and the present disclosure is not limited to the following embodiments.
[0033] (Prior art beam resolution improvement processing) The problems and an overview of the beam resolution improvement process of the conventional technology are shown in Figures 1 and 2. In the left columns of Figures 1 and 2, a weather radar device is equipped with an antenna A that rotates in the azimuth direction. Time series data of a weather radar signal at a certain weather observation point is composed of multiple beams B that rotate in the azimuth direction. In the right columns of Figures 1 and 2, the signal-to-noise ratio of the time series data of the weather radar signal at the weather observation point is improved by data averaging (coherent integration).
[0034] In the right column of Figure 1, the time series data of the weather radar signal at the weather observation point is only applied with a rectangular window function. Therefore, the weights of beam B near both ends are equal to the weights of beam B other than both ends, and the effective beam width of beam B' is significantly deteriorated compared to the original beam width of beam B, so the effect of improving the azimuth beam resolution cannot be obtained.
[0035] In the right column of Figure 2, a window function such as a Hamming window is applied to the time series data of the weather radar signal at the weather observation point. Therefore, the weighting of beam B near both ends is smaller than the weighting of beam B other than the ends, and the effective beam width of beam B" does not deteriorate much compared to the original beam width of beam B, resulting in an improvement in the azimuth beam resolution.
[0036] (Ground surface clutter removal process of the present disclosure) The configuration of the weather radar signal processing device of the present disclosure is shown in Fig. 3. The weather radar signal processing device R includes a ground surface clutter removal device 1, a rainfall signal recovery device 2, and a discontinuity extraction device 3. The weather radar signal processing device R can be realized by installing the weather radar signal processing programs shown in Figs. 4, 7, and 11 in a computer.
[0037] The ground surface clutter removal device 1 (described in FIG. 4 and subsequent figures) includes an autocorrelation matrix calculation unit 11, a ground surface clutter distribution calculation unit 12, a filtering characteristic calculation / storage unit 13, and a ground surface clutter removal unit 14. The rain signal recovery device 2 (described in FIG. 4 and subsequent figures) includes a rain signal distribution calculation unit 21, an autocorrelation matrix calculation unit 22, a filtering processing unit 23, a subtraction unit 24, an addition unit 25, and a rain signal convergence determination unit 26. The discontinuity extraction device 3 (described in FIG. 10 and subsequent figures) includes a rain signal distribution acquisition unit 31, a received power extraction unit 32, a Doppler velocity average extraction unit 33, a Doppler velocity width extraction unit 34, a discontinuity extraction unit 35, and a filtering characteristic change unit 36.
[0038] The procedure and overview of the ground surface clutter removal processing of the present disclosure are shown in FIGS.
[0039] The ground surface clutter removal device 1 performs a time domain filtering process to remove ground surface clutter components from time series data of a weather radar signal to which a window function has been applied (see the right column of FIG. 2) and extract a rainfall signal component, and sets the Doppler velocity width of the ground surface clutter components to be set to a variable width depending on the weather observation point.
[0040] The ground surface clutter removal device 1 inputs the time-series radar signal after the application of the window function at each meteorological observation point and before the removal of ground surface clutter (step S1). As shown in the upper left column of Fig. 5, the Doppler velocity spectrum at the input stage of step S1 includes ground surface clutter components with a Doppler velocity near 0, rain signal components with a finite value of Doppler velocity, and noise components.
[0041] The autocorrelation matrix calculation unit 11 calculates an autocorrelation matrix for the time-series radar signal of step S1 (step S2). The ground surface clutter distribution calculation unit 12 calculates the ground surface clutter distribution for the time-series radar signal of step S1 (step S3). Specifically, the ground surface clutter distribution calculation unit 12 Fourier transforms the time-series radar signal of step S1 into a Doppler velocity spectrum, and calculates the reception intensity and Doppler velocity width of the ground surface clutter distribution.
[0042] The filtering characteristic calculation / storage unit 13 calculates a filtering matrix according to the reception intensity and Doppler velocity width of the ground surface clutter distribution in step S3 (step S4). The ground surface clutter removal unit 14 multiplies the autocorrelation matrix in step S2 by the filtering matrix in step S4 (step S5). As shown in the left middle column of Fig. 5, the Doppler velocity spectrum at the output stage of step S5 includes some rain signal components with finite Doppler velocities and noise components, and some rain signal components near the Doppler velocity of 0 are removed.
[0043] The rain signal recovery device 2 recovers the rain signal components removed by the ground surface clutter removal device 1 together with the ground surface clutter components by calculating the intensity distribution of the received power versus Doppler velocity of the rain signal components extracted by the ground surface clutter removal device 1 based on statistical properties.
[0044] The rain signal distribution calculation unit 21 calculates the rain signal distribution based on the output result of step S5 (step S6). Specifically, the rain signal distribution calculation unit 21 calculates the reception intensity, Doppler velocity average, and Doppler velocity width of the rain signal distribution based on the output result of step S5 being the autocorrelation matrix of the time-series radar signal after removing the ground surface clutter.
[0045] The rain signal distribution calculation unit 21 generates a time series radar signal based on the rain signal distribution of step S6 (step S7). Specifically, the rain signal distribution calculation unit 21 generates a time series radar signal based on the fact that a Gaussian model can be applied to the rain signal distribution of step S6. As shown in the lower left column of FIG. 5, the Doppler velocity spectrum at the output stage of step S7 includes rain signal components near a Doppler velocity of 0, rain signal components with a finite value of the Doppler velocity, and noise components. The autocorrelation matrix calculation unit 22 calculates an autocorrelation matrix for the time series radar signal of step S7 (step S8).
[0046] The filtering processor 23 multiplies the autocorrelation matrix of step S8 by the filtering matrix of step S4 (step S9). As shown in the lower right column of Fig. 5, the Doppler velocity spectrum at the output stage of step S9 contains some rain signal components where the Doppler velocity is a finite value and noise components, and some rain signal components near the Doppler velocity of 0 are removed. That is, step S9 is a stage for preparing the recovery of the rain signal components.
[0047] The subtraction unit 24 subtracts the output result of step S9 from the autocorrelation matrix of step S8 (step S10). As shown in the middle right column of Fig. 5, the Doppler velocity spectrum at the output stage of step S10 includes some rain signal components near the Doppler velocity of 0 and noise components, and some rain signal components where the Doppler velocity is a finite value are removed. In other words, step S10 is a stage for calculating the extent to which the rain signal components are to be restored.
[0048] The adder 25 adds the output result of step S10 to the output result of step S5 (step S11). As shown in the upper right column of FIG. 5, the Doppler velocity spectrum at the output stage of step S11 includes rain signal components with a Doppler velocity close to 0, rain signal components with a finite value of the Doppler velocity, and noise components. That is, step S11 is a stage for realizing recovery of the rain signal components. However, since the rain signal components are not necessarily fully recovered in step S11, the processes of steps S12 to S14 are executed.
[0049] The rain signal distribution calculation unit 21 calculates the rain signal distribution based on the output result of step S11 (step S12). Then, the rain signal distribution calculation unit 21 generates a time-series radar signal based on the rain signal distribution of step S12 (step S13).
[0050] When the rain signal convergence determination unit 26 determines that the rain signal distribution has not converged in step S12 (NO in step S14), it determines that the rain signal component has not been sufficiently recovered. Then, the rain signal recovery device 2 repeats the processes of steps S8 to S14.
[0051] When the rain signal convergence determination unit 26 determines that the rain signal distribution has converged in step S12 (YES in step S14), it determines that the rain signal component has been sufficiently recovered. Then, the rain signal recovery device 2 outputs the time-series radar signal (the time-series radar signal in step S13) after applying the window function at each meteorological observation point and removing the ground surface clutter (step S15).
[0052] In this embodiment, the ground clutter components are removed by multiplying the autocorrelation matrix of the time series radar signal by a filtering matrix. As a modified example, the ground clutter components may be removed by performing FIR filtering or IIR filtering on the time series radar signal. In this embodiment, the rain signal components are restored by applying a Gaussian model to the rain signal distribution and then multiplying the autocorrelation matrix of the time series radar signal after the ground clutter has been removed by a filtering matrix. As a modified example, the rain signal components may be restored by only applying a Gaussian model to the rain signal distribution, without multiplying the autocorrelation matrix of the time series radar signal after the ground clutter has been removed by a filtering matrix.
[0053] (Filtering characteristic calculation process of the present disclosure) A method in which the filtering characteristic calculation / storage unit 13 calculates the filtering matrix in accordance with the reception intensity and Doppler velocity width of the ground surface clutter distribution will be described below.
[0054] Doppler velocity spectrum S of ground clutter components C (v) is expressed as number 1. Here, P C is the ground clutter intensity, and σ C is the ground clutter velocity width.
number
number
number
[0055] Autocorrelation function R of time series radar signals of surface clutter components C (τ) (where τ is the delay time) is expressed as Equation 4 for wavelength λ.
number
number
number
number
[0056] Autocorrelation function R of time series radar signals of surface clutter components C Autocorrelation matrix R of (τ) C Here, m is the autocorrelation function R of the time series radar signal. C (τ), and * denotes the complex conjugate.
number
number
number
number
[0057] Multiplying the autocorrelation matrix R of the autocorrelation function R(τ) of the time-series radar signal, which is a combination of the ground clutter component, the precipitation signal component, and the noise component, by the filtering matrix A gives Equation 12. Here, R' is the multiplication result, H represents the complex conjugate transpose, and I m represents an m-row, m-column unit matrix.
number
[0058] When the autocorrelation function R(τ) of the time series radar signal does not include the rain signal component, the condition of Equation 13 is necessary for the ground clutter component in the multiplication result R' to be reduced to the noise level.
number
number
[0059] The problem with the filtering characteristic calculation process of the present disclosure is shown in FIG. 6. In the upper part of FIG. 6, a window function such as a Hamming window is not applied to the time series data of the weather radar signal, and the filtering process in the time domain is not performed. Then, the ground clutter components cannot be appropriately removed. In the middle part of FIG. 6, a window function such as a Hamming window is not applied to the time series data of the weather radar signal, and only a rectangular window function is applied to the filtering process in the time domain. Then, the ground clutter components can be appropriately removed. In the lower part of FIG. 6, a window function such as a Hamming window is applied to the time series data of the weather radar signal, and only a rectangular window function is applied to the filtering process in the time domain. Then, the ground clutter components cannot be appropriately removed.
[0060] (Filtering characteristic storage process of the present disclosure) In the above, a window function is applied to the time series data of the weather radar signal, but the same window function is not applied to the filtering process in the time domain. If a window function is applied to the time series data of the weather radar signal, the statistical properties of the time series data of the weather radar signal, such as Gaussian, change. Therefore, it is not possible to appropriately remove the ground clutter components, and it is not possible to achieve both the effect of removing the ground clutter components and the effect of improving the azimuth beam resolution.
[0061] In the following, a window function is applied to the time series data of the weather radar signal, and the same window function is applied to the filtering process in the time domain. Here, applying a window function to the time series data of the weather radar signal changes the statistical properties of the time series data of the weather radar signal, such as Gaussian. However, it is possible to appropriately remove the ground clutter components, and it is possible to achieve both the effect of removing the ground clutter components and the effect of improving the azimuth beam resolution.
[0062] In the above, the filtering characteristic calculation / storage unit 13 calculates various filtering characteristics in the time domain according to various reception intensities and Doppler velocity widths of ground surface clutter components during meteorological observation. Since the calculation of the filtering matrix A includes the calculation of an inverse matrix as described above in Equation 14, the calculation processing load increases during meteorological observation.
[0063] In the following, the filtering characteristic calculation / storage unit 13 stores various filtering characteristics in the time domain according to various reception intensities and Doppler velocity widths of ground surface clutter components prior to meteorological observation. Then, even when the calculation of the filtering matrix A includes the calculation of the inverse matrix as described above in Equation 14, the calculation processing load during meteorological observation is reduced.
[0064] The procedure of the filtering characteristic storage process of the present disclosure is shown in Fig. 7. The processes of steps S21 to S are processes of the filtering characteristic calculation / storage unit 13.
[0065] Doppler velocity spectrum S of various ground clutter components to be removed C (v) is set as shown in Equation 1 (step S21). The clutter strength P C and clutter velocity width σ C (Step S22). Here, the clutter intensity P C and clutter velocity width σ CThe narrower the interval between the filtering matrices A, the easier it is to select the optimal filtering matrix A from among various filtering matrices A, and the higher the accuracy of ground clutter removal.
[0066] Autocorrelation function R of time series radar signals with various ground clutter components to be removed C (τ) is calculated as shown in Equation 4 (step S23). The autocorrelation function R C For (τ), the autocorrelation matrix R after applying the window function C is calculated as in Equation 15 (step S24). Here, w1, . . . , w m is a window function, such as a Hamming window, that is applied to the filtering matrix A as well as the time series radar signal.
number
[0067] Autocorrelation function R of the time series radar signal with the expected noise component N (τ) is set as shown in Equation 6 (step S25). The noise intensity σ after applying the window function to the expected noise component is N 2 is set as in Equation 16 (step S26). Here, the noise intensity σ N 2 The higher the accuracy of setting w1,...,w m is a window function, such as a Hamming window, that is applied to the filtering matrix A as well as the time series radar signal.
number
[0068] A filtering matrix A after application of a window function that removes various ground surface clutter components is calculated as shown in Equation 14 (step S27). The filtering matrix A after application of a window function that removes various ground surface clutter components is stored (step S28). An optimal filtering matrix A is selected according to the reception intensity and Doppler velocity width of the ground surface clutter distribution in step S3 (step S4). Here, a method for selecting the optimal filtering matrix A will be described.
[0069] If a window function is applied to the time series data of the weather radar signal, the reception strength and Doppler velocity width of the ground clutter components are reduced. Therefore, prior to meteorological observation, a window function is applied to the time series data of the weather radar signal to calculate the reduction in the reception strength and Doppler velocity width of the ground clutter components. Then, in the meteorological observation, the reduction in the reception strength and Doppler velocity width of the ground clutter components are added to the reception strength and Doppler velocity width of the ground clutter components after the application of the window function to calculate the reception strength and Doppler velocity width of the ground clutter components before the application of the window function. Then, in the meteorological observation, an optimal filtering matrix A can be selected according to the reception strength and Doppler velocity width of the ground clutter components before the application of the window function.
[0070] If a window function is applied to the time series data of the weather radar signal, the received strength of the noise components is reduced. Therefore, prior to meteorological observation, a window function is applied to the time series data of the weather radar signal to calculate the amount of reduction in the received strength of the noise components. Then, in the meteorological observation, the amount of reduction in the received strength of the noise components is added to the received strength of the noise components after the window function is applied to calculate the received strength of the noise components before the window function is applied. Then, in the meteorological observation, the optimal filtering matrix A can be selected according to the received strength of the noise components before the window function is applied.
[0071] The effect of the filtering characteristic calculation process of the present disclosure is shown in FIG. 8. In the upper part of FIG. 8, a window function such as a Hamming window is not applied to the time series data of the weather radar signal, and the filtering process in the time domain is not performed. As a result, the ground clutter components cannot be appropriately removed. In the middle part of FIG. 8, a window function such as a Hamming window is not applied to the time series data of the weather radar signal, and only a rectangular window function is applied to the filtering process in the time domain. As a result, the ground clutter components can be appropriately removed. In the lower part of FIG. 8, a window function such as a Hamming window is applied to the time series data of the weather radar signal, and a window function such as a Hamming window is applied to the filtering process in the time domain. As a result, the ground clutter components can be appropriately removed.
[0072] The effect of the filtering characteristic calculation process of the present disclosure is also shown in Fig. 9. In the upper part of Fig. 9, a window function such as a Hamming window is not applied to the time series data of the weather radar signal, and only a rectangular window function is applied to the time domain filtering process. In the middle part of Fig. 9, a window function such as a Hamming window is applied to the time series data of the weather radar signal, and a window function such as a Hamming window is applied to the time domain filtering process. In the lower part of Fig. 9, the upper and middle parts of Fig. 9 are compared, and in the upper part of Fig. 9 (super-resolution OFF), the azimuth beam resolution cannot be improved, but in the middle part of Fig. 9 (super-resolution ON), the azimuth beam resolution can be improved.
[0073] (Ground surface clutter removal range setting process of the present disclosure) In the above, when calculating the filtering matrix A, the Doppler velocity width σ C is set to a variable width for all weather observation points, rather than being set to a fixed width only.
[0074] Therefore, the Doppler velocity width σ C In order to remove the small ground clutter components, the variable Doppler velocity width σ C By appropriately narrowing, it is possible to reduce the possibility that components other than the ground clutter are excessively removed. CIn order to remove the large tree clutter components, the variable Doppler velocity width σ C By appropriately widening, the possibility that the tree clutter component will not be completely removed can be reduced.
[0075] In the following, when calculating the filtering matrix A, the Doppler velocity width σ C (1) In the initial processing, the variable Doppler velocity width σ is set to a fixed width only for all meteorological observation points, and (2) in the post-processing, the variable width is changed to a variable width only for a portion of meteorological observation points where it is difficult to remove the ground clutter components or extract the rain signal components in the initial processing, and the fixed width is maintained for most of the meteorological observation points where it is possible to remove the ground clutter components or extract the rain signal components even in the initial processing. C This can reduce the burden of setting up the device.
[0076] An outline of the ground surface clutter removal range setting process of the present disclosure is shown in FIG. P The rain parameters including the rain velocity range and the rain velocity width are considered to be spatially continuous. A rain point P1 in the rain area P is located in a direction shifted from the front direction of the weather radar signal processing device R, and the Doppler velocity v D is a positive finite value. The rainfall point P2 in the rainfall area P is located almost in front of the weather radar signal processing device R, and the Doppler velocity v D is almost zero.
[0077] First, the Doppler velocity width σ P is not too narrow, and the Doppler velocity width σ of the ground clutter component is C Consider the case where is not too wide (second row from the top of Figure 10).
[0078] For rainfall point P1, the finite value of the Doppler velocity v D In extracting the rain signal component whose Doppler velocity is the mean, the Doppler velocity width σ Pis not too narrow, so that if the fixed removal range of the ground clutter components is appropriately set, there is a low possibility that the rain signal components will be excessively removed or excessively left, and there is a high possibility that the rain signal components can be recovered. Therefore, the fixed removal range of the ground clutter components is maintained as it is.
[0079] For rainfall point P2, the Doppler velocity v near 0 D In extracting the rain signal component whose Doppler velocity is the mean, the Doppler velocity width σ P is not too narrow, so that if the fixed removal range of the ground clutter components is appropriately set, there is a low possibility that the rain signal components will be excessively removed or excessively left, and there is a high possibility that the rain signal components can be recovered. Therefore, the fixed removal range of the ground clutter components is maintained as it is.
[0080] From the rainfall point P1 to the rainfall point P2, the rainfall parameters including the received intensity of the rainfall signal component, the Doppler velocity average, and the Doppler velocity width are spatially continuous.
[0081] Next, the Doppler velocity width σ of the ground clutter component C is not too wide, but the Doppler velocity width σ of the rain signal component P Consider the case where is too narrow (third row from the top in Figure 10).
[0082] For rainfall point P1, the finite value of the Doppler velocity v D In extracting the rain signal component whose Doppler velocity is the mean, the Doppler velocity width σ P However, if the fixed removal range of the ground clutter components is appropriately set, the possibility that the rain signal components are excessively removed or excessively left is low, and it is highly likely that the rain signal components can be recovered. Therefore, the fixed removal range of the ground clutter components is maintained as it is.
[0083] For rainfall point P2, the Doppler velocity v near 0 DIn extracting the rain signal component whose Doppler velocity is the mean, the Doppler velocity width σ P Since the fixed removal range of the ground clutter components is too narrow, even if the fixed removal range of the ground clutter components is appropriately set, there is a high possibility that the rain signal components will be removed in excess, and the rain signal components will become unrecoverable. Therefore, it is desirable to change the removal range of the ground clutter components to a variable width narrower than the fixed width.
[0084] From the rainfall point P1 to the rainfall point P2, the rainfall parameters including the reception strength of the rainfall signal component, the Doppler velocity average, and the Doppler velocity range are spatially discontinuous. In particular, since the rainfall signal component becomes unrecoverable, the reception strength of the rainfall signal component is likely to become spatially discontinuous. Furthermore, if the Gaussian fitting in step S7 is not appropriate, the Doppler velocity average and the Doppler velocity range of the rainfall signal component will also become spatially discontinuous.
[0085] Next, the Doppler velocity width σ P is not too narrow, but the Doppler velocity width σ of the ground clutter component C Consider the case where is too wide (fourth row from the top in Figure 10).
[0086] For rainfall point P1, the finite value of the Doppler velocity v D In extracting the rain signal component, whose Doppler velocity is the mean, the Doppler velocity width σ C However, if the fixed removal range of the ground clutter components is appropriately set, the possibility that the rain signal components are excessively removed or excessively left is low, and it is highly likely that the rain signal components can be recovered. Therefore, the fixed removal range of the ground clutter components is maintained as it is.
[0087] For rainfall point P2, the Doppler velocity v near 0 D In extracting the rain signal component, whose Doppler velocity is the mean, the Doppler velocity width σ Cis too wide, so even if the fixed removal range of the ground clutter components is appropriately set, there is a high possibility that the remaining ground clutter components will be extracted as spurious rain signal components, and the rain signal components will be over-recovered. Therefore, it is desirable to change the removal range of the ground clutter components to a variable width that is wider than the fixed width.
[0088] From the rainfall point P1 to the rainfall point P2, the rainfall parameters including the reception strength of the rainfall signal component, the Doppler velocity average, and the Doppler velocity range are spatially discontinuous. In particular, the rainfall signal component is over-recovered, so the reception strength of the rainfall signal component is likely to become spatially discontinuous. If the Gaussian fitting in step S7 is not appropriate, the Doppler velocity average and the Doppler velocity range of the rainfall signal component will also become spatially discontinuous.
[0089] Conversely, a meteorological observation point (discontinuous observation point) where the parameters of a rainfall signal component that is actually continuous are erroneously discontinuous is regarded as a meteorological observation point where the rainfall signal component is unrecoverable or over-recovered. Therefore, it is possible to distinguish between meteorological observation points where the rainfall signal component is unrecoverable and meteorological observation points where the rainfall signal component is actually small. It is also possible to distinguish between meteorological observation points where the rainfall signal component is over-recovered and meteorological observation points where the rainfall signal component is actually large.
[0090] FIG. 11 shows the procedure of the ground surface clutter removal range setting process according to the present disclosure.
[0091] The filtering characteristic calculation / storage unit 13 calculates the Doppler velocity width σ of the ground surface clutter component to be set in the filtering process in the time domain. C is set to a fixed width regardless of the meteorological observation point as an initial process. Specifically, the filtering characteristic calculation / storage unit 13 calculates the Doppler velocity width σ of the ground surface clutter component based on the antenna beam width and scan speed and the characteristics of the ground surface clutter component. C is set as a fixed width. Then, the ground surface clutter remover 1 and the rain signal recovery device 2 execute the processes of steps S1 to S15 (step S31).
[0092] The rain signal distribution acquisition unit 31 acquires the rain signal distribution in step S12 at each meteorological observation point (step S32). The received power extraction unit 32, the Doppler velocity average extraction unit 33, and the Doppler velocity range extraction unit 34 extract the received power, the Doppler velocity average, and the Doppler velocity range from the rain signal distribution in step S32 at each meteorological observation point (step S33).
[0093] The discontinuity extraction unit 35 extracts discontinuous observation points where at least one of the received power, Doppler velocity average, and Doppler velocity width of the rain signal component recovered by the rain signal recovery device 2 has discontinuity within the meteorological observation space (step S34).
[0094] The filtering characteristic change unit 36 changes the Doppler velocity width σ of the ground surface clutter components to be set in the time domain filtering process. C to a variable width different from the fixed width at the discontinuous observation points extracted by the discontinuous point extracting unit 35 as post-processing (YES in step S34, step S35). Then, the ground surface clutter removing device 1 and the rain signal recovery device 2 execute the processes of steps S1 to S15 again (step S35). Here, steps S34 and S35 may be executed repeatedly until there are no discontinuous observation points in step S34.
[0095] Here, the specific procedures of steps S34 and S35 will be described.
[0096] The discontinuous point extraction unit 35 determines whether the sign of the Doppler velocity average at each meteorological observation point is inverted compared to the surrounding meteorological observation points (step S41). In other words, the discontinuous point extraction unit 35 searches for meteorological observation points whose Doppler velocity average is close to 0.
[0097] For meteorological observation points whose Doppler velocity average reverses sign compared to the surrounding meteorological observation points (YES in step S42), there is a possibility that the rain signal component has been excessively removed or excessively remained, so the process proceeds to step S43. For meteorological observation points whose Doppler velocity average does not reverse sign compared to the surrounding meteorological observation points (NO in step S42), there is a low possibility that the rain signal component has been excessively removed or excessively remained, so the process ends.
[0098] The discontinuous point extraction unit 35 determines whether the received power or the power difference before and after recovery at each meteorological observation point is smaller or larger than a predetermined difference compared to the surrounding meteorological observation points (step S43). In other words, the discontinuous point extraction unit 35 searches for meteorological observation points where the rain signal component has been excessively removed or excessively remained, and extracts the meteorological observation points as discontinuous observation points.
[0099] For a meteorological observation point whose received power or the power difference before and after recovery is smaller than a predetermined difference or more compared to the surrounding meteorological observation points ("small" in step S44), the filtering characteristic change unit 36 sets the Doppler velocity width σ C to a variable width narrower than the fixed width as post-processing (step S45). Then, the ground surface clutter remover 1 and the rain signal recovery device 2 execute the processes of steps S1 to S15 again (step S45). Here, steps S34 to S35 and S41 to S46 may be repeatedly executed until there are no discontinuous observation points in steps S34 and S44.
[0100] For a meteorological observation point whose received power or the power difference before and after recovery is greater than or equal to a predetermined difference compared to surrounding meteorological observation points ("Large" in step S44), the filtering characteristic change unit 36 sets the Doppler velocity width σ Cto a variable width wider than the fixed width as post-processing (step S46). Then, the ground surface clutter remover 1 and the rain signal recovery device 2 execute the processes of steps S1 to S15 again (step S46). Here, steps S34 to S35 and S41 to S46 may be repeatedly executed until there are no discontinuous observation points in steps S34 and S44.
[0101] For meteorological observation points whose received power or the power difference before and after recovery is neither smaller nor larger than a specified difference compared to surrounding meteorological observation points ("neither" in step S44), the processing is terminated because there is a low possibility of excessive removal or excessive remaining of rain signal components.
[0102] The window function applied to the time series data of the weather radar signal can be set according to the rotation speed of antenna A and the size of the desired target prior to meteorological observation, and can be fixed rather than made variable even if discontinuous observation points of the rain signal component are extracted during meteorological observation. [Industrial Applicability]
[0103] The weather radar signal processing device and weather radar signal processing program disclosed herein, when performing time-domain filtering on a weather radar signal, (1) achieves both the effect of removing ground surface clutter components and the effect of improving azimuth beam resolution, (2) reduces the possibility that ground surface clutter will be removed in excess or incompletely, and reduces the possibility that precipitation signal components will become unrecoverable or over-recovered, and (3) reduces the computational processing burden during meteorological observation. [Explanation of symbols]
[0104] A: Antenna B, B', B”: Beam R: Weather radar signal processing device 1: Ground clutter removal device 2: Rain signal recovery device 3: Discontinuity extraction device 11: Autocorrelation matrix calculation section 12: Ground surface clutter distribution calculation part 13: Filtering characteristic calculation / storage section 14: Ground surface clutter removal section 21: Rainfall signal distribution calculation section 22: Autocorrelation matrix calculation unit 23: Filtering processing unit 24: Subtraction section 25: Addition section 26: Rainfall signal convergence judgement unit 31: Rainfall signal distribution acquisition section 32: Received power extraction unit 33: Doppler velocity average extraction unit 34: Doppler velocity width extraction unit 35: Discontinuity extraction section 36: Filtering characteristic change section P:Rainfall area P1, P2: Rainfall point
Claims
1. a weather radar signal acquisition unit that acquires time series data of a weather radar signal to which a window function has been applied from a weather radar device having an antenna that rotates in an azimuth angle direction; a ground clutter removal unit that performs a filtering process in the time domain to which the window function is applied to the time series data of the weather radar signal when removing ground clutter components from the time series data of the weather radar signal to which the window function is applied and extracting a rainfall signal component; a rain signal recovery unit that recovers the rain signal components removed by the ground clutter removal unit together with the ground clutter components by calculating, based on a statistical property, an intensity distribution of received power with respect to the Doppler velocity of the rain signal components extracted by the ground clutter removal unit; A weather radar signal processing device comprising:
2. The ground surface clutter removal unit sets a Doppler velocity width of the ground surface clutter component to be set in a time domain filtering process to which the window function is applied to the time series data of the weather radar signal, to a variable width depending on a meteorological observation point.
2. A weather radar signal processing device according to claim 1 .
3. the ground clutter removal unit sets a Doppler velocity width of the ground clutter components to be set to a fixed width as an initial process, regardless of the weather observation point, in a time domain filtering process to which the window function is applied to the time series data of the weather radar signal, The weather radar signal processing device further includes a discontinuity extraction unit that extracts discontinuous observation points in which at least one of the received power, the Doppler velocity average, and the Doppler velocity width of the rain signal component recovered by the rain signal recovery unit has discontinuity in a meteorological observation space, The ground surface clutter removal unit changes a Doppler velocity width of the ground surface clutter component to be set in a time domain filtering process to which the window function applied to the time series data of the weather radar signal is applied, to the variable width different from the fixed width at the discontinuous observation points extracted by the discontinuous point extraction unit as a post-processing.
3. The weather radar signal processing device according to claim 2, wherein the weather radar signal processing device is a signal processing device.
4. The discontinuous point extraction unit extracts, as the discontinuous observation points, meteorological observation points where the Doppler velocity average of the rain signal component recovered by the rain signal recovery unit has an inverted sign compared to surrounding meteorological observation points, and where the received power of the rain signal component recovered by the rain signal recovery unit or the power difference before and after recovery is smaller than a predetermined difference or more compared to surrounding meteorological observation points, The ground surface clutter removal unit changes the Doppler velocity width of the ground surface clutter component to be set in a time domain filtering process to which the window function applied to the time series data of the weather radar signal is applied, to the variable width narrower than the fixed width at the discontinuous observation points extracted by the discontinuous point extraction unit as a post-processing.
4. The weather radar signal processing device according to claim 3, wherein the weather radar signal processing device is a weather radar signal processing device.
5. The discontinuous point extraction unit extracts, as the discontinuous observation points, meteorological observation points where the Doppler velocity average of the rain signal component recovered by the rain signal recovery unit has an inverted sign compared to surrounding meteorological observation points, and where the received power of the rain signal component recovered by the rain signal recovery unit or the power difference before and after recovery is greater than or equal to a predetermined difference compared to surrounding meteorological observation points, The ground surface clutter removal unit changes the Doppler velocity width of the ground surface clutter component to be set in a time domain filtering process to which the window function applied to the time series data of the weather radar signal is applied, to the variable width wider than the fixed width at the discontinuous observation points extracted by the discontinuous point extraction unit as a post-processing.
4. The weather radar signal processing device according to claim 3, wherein the weather radar signal processing device is a weather radar signal processing device.
6. The ground surface clutter removal unit stores, prior to meteorological observation, various filtering characteristics in the time domain to which the window function is applied to the time series data of the weather radar signal in accordance with various reception intensities and Doppler velocity widths of ground surface clutter components.
6. A weather radar signal processing device according to claim 1, wherein the weather radar signal processing device is a weather radar signal processing device.
7. a weather radar signal acquisition step of acquiring time series data of a weather radar signal to which a window function has been applied from a weather radar device having an antenna that rotates in an azimuth angle direction; a ground clutter removal step of performing a time domain filtering process to which the window function is applied to the time series data of the weather radar signal, in order to remove ground clutter components from the time series data of the weather radar signal to which the window function is applied and extract a rainfall signal component; a rain signal recovery step of recovering the rain signal components removed together with the ground clutter components in the ground clutter removal step by calculating, based on statistical properties, a distribution of intensity of received power with respect to the Doppler velocity of the rain signal components extracted in the ground clutter removal step; A weather radar signal processing program for causing a computer to execute the above steps in sequence.
Citation Information
Patent Citations
Rainfall signal recovery method and device of polarization-Doppler meteorological radar
CN114415184A
Radar signal processing device and method thereof
JP2002131421A
Meteorological radar signal processor and meteorological radar signal processing program
JP2020106471A
System and method for processing data in weather radar
US20080001808A1
System and method for increasing spectral resolution
US20100231439A1
