A staggered pulse repetition interval time-frequency domain dual-polarized radar detection variable estimation method

Abstract

The present application relates to radar signal processing technical field, specifically provide a kind of staggered pulse repetition interval time-frequency domain dual-polarized radar detection variable estimation method, comprising: in frequency domain, the estimation of echo signal-to-noise ratio, reflectivity factor, differential reflectivity of horizontal / vertical polarization based on power spectral density;And the estimation of average Doppler velocity and velocity spectrum width;In time domain, after ground clutter filtering based on odd-even pulse, respectively, odd-even pulse correlation parameter estimation is carried out, so that two detection variables of dual-polarized radar correlation coefficient and differential propagation phase shift of odd-even pulse are obtained respectively, then odd-even combination can obtain the estimation result of accurate correlation coefficient and differential propagation phase shift.The present application realizes the accurate estimation of dual-polarized radar multi-class detection variable, ensures that estimation result is not disturbed by ground clutter interference, simultaneously adapts single pulse repetition interval and staggered pulse repetition interval two working modes, improves the reliability and applicability of radar detection.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing technology, and in particular to a method for estimating detection variables in a time-frequency domain dual-polarization radar with staggered pulse repetition intervals. Background Technology

[0002] Traditional radar signal processing relies solely on either time-domain correlation moments or frequency-domain power spectral density methods, which presents numerous technical bottlenecks.

[0003] Time-domain processing defects: IIR filters lose meteorological signals without compensation. Under uneven pulse repetition intervals, the low sampling rate limits the filter order, leading to performance degradation. There is also a blind speed effect (useful meteorological signals are misjudged as ground clutter).

[0004] Limitations of frequency domain processing: Under uneven pulse repetition intervals, low sampling rates do not satisfy the sampling theorem, leading to sampling errors and making it impossible to perform effective parameter statistics.

[0005] The core problem is that a single pulse repetition interval cannot simultaneously achieve both "unambiguous distance" and "unambiguous velocity," forcing a choice between the two, which makes it difficult to meet actual detection requirements. Summary of the Invention

[0006] This invention provides a time-frequency domain method for estimating detection variables in dual-polarized radar with staggered pulse repetition intervals. It overcomes the inherent contradictions in traditional radar signal processing, such as ground clutter interference, blind velocity effect, insufficient sampling rate, and the need for unambiguous range and velocity. This method achieves accurate estimation of multiple detection variables in dual-polarized radar, ensuring that the estimation results are not affected by ground clutter. It is also compatible with both single-pulse repetition interval and staggered pulse repetition interval working modes, thus improving the reliability and applicability of radar detection.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A method for estimating detection variables in a time-frequency domain dual-polarization radar with staggered pulse repetition intervals, comprising:

[0009] The horizontally polarized complex baseband signal and the vertically polarized complex baseband signal output by the radar digital receiver are acquired. Discrete Fourier transforms are performed on the horizontally polarized complex baseband signal and the vertically polarized complex baseband signal respectively to obtain the horizontally polarized power spectral density and the vertically polarized power spectral density.

[0010] Based on the horizontal polarization power spectral density and the vertical polarization power spectral density, the horizontal polarization noise power and the vertical polarization noise power are determined, and then the estimated values ​​of the horizontal polarization noise power and the estimated values ​​of the vertical polarization noise power are calculated.

[0011] Using the horizontal polarization power spectral density and the vertical polarization power spectral density, the average echo power of horizontal polarization and vertical polarization at each distance gate is calculated respectively. The signal-to-noise ratio of horizontal polarization echo and the signal-to-noise ratio of vertical polarization echo are obtained by combining the estimated values ​​of horizontal polarization noise power and vertical polarization noise power.

[0012] Based on the average echo power of horizontal and vertical polarization at each range gate, the signal-to-noise ratio of horizontal polarization echo, and the signal-to-noise ratio of vertical polarization echo, the radar reflectivity factor and differential reflectivity are calculated. At the same time, the average Doppler velocity and velocity spectral width are estimated based on the power spectral density of horizontal polarization and the power spectral density of vertical polarization.

[0013] The horizontally polarized complex baseband signal and the vertically polarized complex baseband signal are respectively subjected to odd-even pulse high-pass filtering to remove ground clutter interference and obtain the filtered odd-even pulses; the cross-correlation estimation and autocorrelation estimation of the filtered odd-even pulses are respectively performed to obtain the cross-correlation results and autocorrelation results of the odd-even pulses.

[0014] The cross-correlation results of the odd and even pulses are normalized with the corresponding autocorrelation results, and the absolute values ​​are taken and averaged to obtain the correlation coefficient.

[0015] The argument of the odd-even pulse cross-correlation results is calculated and averaged to obtain the initial differential propagation phase shift;

[0016] The initial differential propagation phase shift is processed to obtain the corrected differential propagation phase shift;

[0017] The radar reflectivity factor, differential reflectivity, average Doppler velocity, and velocity spectrum width are smoothed by moving average. The correlation coefficient and the corrected differential propagation phase shift are smoothed by moving median filtering. Finally, the estimation results of all radar detection variables are obtained.

[0018] In this specification, the odd-even pulse high-pass filtering is implemented using one of the following two methods: DC removal processing: The average signal values ​​of the odd pulse group and even pulse group in the horizontally polarized and vertically polarized complex baseband signals are calculated respectively. The average value is taken as the DC component of the corresponding pulse group. Then, the DC component of the pulse group to which each signal sample belongs is subtracted, and the AC component is retained to remove the main component of ground clutter; IIR high-pass filtering: A 6th-order elliptic filter is used. The stopband speed of the filter is set to 0.7 m / s and the passband speed is set to 1 m / s. A difference equation is constructed through preset autoregressive coefficients and moving average coefficients to filter the odd and even pulse groups respectively to completely suppress ground clutter.

[0019] In this specification, the specific process for determining the horizontal and vertical polarization noise power is as follows: The horizontal and vertical polarization power spectral densities of each distance gate are sorted from low to high to obtain the corresponding power spectral density sequences; the cumulative power sum of each sequential point in each sorted power spectral density sequence is calculated to form a power sum distribution map; based on the empirical rule that noise power distribution falls within the 5% to 40% power sum range, the starting point where the power jump in the power sum distribution map exceeds a preset threshold is identified, and the cumulative power sum corresponding to this starting point is the horizontal or vertical polarization noise power of the corresponding distance gate; the average horizontal and vertical polarization noise power values ​​of multiple distance gates and different radial directions are statistically analyzed and used as the estimated values ​​of the horizontal and vertical polarization noise power, respectively.

[0020] In this specification, after performing discrete Fourier transform on the horizontally polarized complex baseband signal and the vertically polarized complex baseband signal, a ground clutter suppression step is also included: determining the digital frequency half-bandwidth corresponding to the ground clutter, and the power spectral density of the target echo signal adjacent to the low-frequency end and high-frequency end of the ground clutter; calculating the rate of change of the signal power spectral density based on the target echo signal power spectral density; and using the linear interpolation result corresponding to the rate of change to replace the value of the part of the power spectral density affected by ground clutter, to obtain the horizontally polarized power spectral density and the vertically polarized power spectral density after clutter removal.

[0021] In this specification, the cross-correlation estimation is performed separately according to the grouping method of odd pulse short interval and even pulse long interval; the autocorrelation estimation is performed separately according to the polarization type of horizontal polarization and vertical polarization, combined with the grouping method of odd pulse short interval and even pulse long interval, and the autocorrelation estimation result is used to characterize the echo power of the corresponding pulse group; the absolute value is taken after normalization processing to limit the correlation coefficient to the real number range of 0 to 1, so as to avoid the small imaginary components generated by complex number calculation from affecting the estimation result.

[0022] In this specification, the continuous processing of the initial differential propagation phase shift specifically includes: eliminating transient jumps: detecting the jump amplitude of the initial differential propagation phase shift between adjacent range gates; if the jump amplitude exceeds the phase jump perturbation threshold but is less than the maximum phase jump range threshold, then the initial differential propagation phase shift of the current range gate is corrected to the value of the previous range gate; eliminating out-of-range jumps: if the jump amplitude exceeds the phase forward folding threshold or the phase reverse folding threshold, the initial differential propagation phase shift of the next range gate is adjusted and the jump amplitude is recalculated, replacing the abnormal measurement with the valid measurement of the adjacent range gate; and normalizing the numerical range: if the initial differential propagation phase shift is greater than 180°, then 360° is subtracted; if it is less than -180°, then 360° is added, ensuring that the corrected differential propagation phase shift is within the range of ±180°.

[0023] In this specification, the number of sliding averages for the sliding average smoothing process is set to 10 distance gates; the processing length of the sliding median filter smoothing process corresponds to the number of distance gates in the distance range of 300m to 500m, in order to eliminate the small fluctuations generated during the processing and ensure that the estimation results are smooth and undistorted.

[0024] In this specification, the Discrete Fourier Transform is adapted to two operating modes: single-pulse repetition interval and staggered-pulse repetition interval. When the radar operates in single-pulse repetition interval mode, Fast Fourier Transform is used to process the horizontally polarized complex baseband signal and the vertically polarized complex baseband signal to directly obtain the horizontally polarized power spectral density and the vertically polarized power spectral density. When the radar operates in staggered-pulse repetition interval mode, staggered Discrete Fourier Transform technology is used to process odd-pulse groups and even-pulse groups separately to ensure the uniformity of the sampling interval, thereby obtaining the horizontally polarized power spectral density and the vertically polarized power spectral density.

[0025] In this specification, when using the staggered discrete Fourier transform technique, the image frequency component removal step is also included: for each distance gate, the position of the strongest power spectral density among the horizontal polarization power spectral density and the vertical polarization power spectral density is found; based on the characteristic that the amplitude of the main frequency region is greater than the amplitude of the image frequency component, the effective frequency domain range is determined with the position of the strongest power spectral density as the benchmark, and the image frequency component generated by zero-padding interpolation is removed to obtain a pure power spectral density.

[0026] In this specification, the calculation of the differential reflectivity also includes attenuation correction, channel error correction, and observation error correction: attenuation correction is performed based on the attenuation coefficient determined by fitting after actual observation; channel error correction describes the power measurement difference between horizontal and vertical polarization within the normal linear dynamic range of the radar using a quadratic fitting curve determined during factory calibration; observation error correction is performed based on the observation error value determined by field calibration to improve the estimation accuracy of differential reflectivity.

[0027] In summary, this invention has at least the following beneficial effects: it completely solves the problem of ground clutter interference, ensuring that all radar detection variables are unaffected by ground clutter; it effectively overcomes the shortcomings of traditional technologies such as blind velocity effect and insufficient sampling rate, reconciling the fundamental contradiction between unambiguous range and unambiguous velocity; it achieves accurate estimation of multiple types of detection variables (amplitude, velocity, cross-correlation class), with smooth and representative results; it adapts to both single-pulse and staggered pulse repetition interval working modes, exhibiting strong compatibility, and is particularly suitable for meteorological detection scenarios of dual-polarization Doppler weather radar; the processing flow is both simple and practical, with no delay in time-domain DC removal processing and more thorough IIR filtering to remove clutter, allowing for flexible selection according to radar application scenarios. Attached Figure Description

[0028] Figure 1This is a schematic diagram of a method for estimating detection variables in a dual-polarization radar in the frequency domain for staggered pulse repetition intervals.

[0029] Figure 2 This is a schematic diagram of the measured horizontal polarization power spectral density under staggered pulse repetition intervals.

[0030] Figure 3 This is a schematic diagram of the measured vertical polarization power spectral density under staggered pulse repetition intervals.

[0031] Figure 4 This is a schematic diagram showing the power spectral density, average Doppler velocity, and velocity spectral width estimation at a range gate in the frequency domain.

[0032] Figure 5 This is a schematic diagram showing the power spectral density sorted from low to high on a distance gate in the frequency domain, as well as the signal-noise distinguishing features.

[0033] Figure 6 This is a schematic diagram of the frequency domain processing of signal-to-noise ratio and differential reflectivity estimation results.

[0034] Figure 7 This is a schematic diagram of the reflectivity factor estimation results processed in the frequency domain.

[0035] Figure 8 This is a schematic diagram showing the actual results of estimating the strongest signal Doppler velocity, average Doppler velocity, and velocity spectral width in the frequency domain using a range-gated scanning method.

[0036] Figure 9 This is a schematic diagram of the signal flow of an IIR filter for suppressing ground clutter under staggered pulse repetition intervals.

[0037] Figure 10 This is a schematic diagram of the frequency domain amplitude response characteristics of an IIR filter for suppressing ground clutter under staggered pulse repetition intervals.

[0038] Figure 11 A schematic diagram of the process for eliminating transient phase shift jumps during differential propagation.

[0039] Figure 12 A schematic diagram of the process for eliminating phase shift jumps that exceed the range during differential propagation.

[0040] Figure 13 This is a schematic diagram of the process for handling differential propagation phase shift within the normal range.

[0041] Figure 14 This is a schematic diagram showing the estimation results of the DC correlation coefficient under the staggered pulse repetition interval.

[0042] Figure 15 A schematic diagram of the differential propagation phase shift result of DC under staggered pulse repetition intervals. Detailed Implementation

[0043] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0044] like Figure 1 As shown, this embodiment provides a method for estimating detection variables in a time-frequency domain dual-polarization radar with staggered pulse repetition intervals, including: acquiring the horizontally polarized complex baseband signal and the vertically polarized complex baseband signal output by a radar digital receiver; performing discrete Fourier transforms on the horizontally polarized complex baseband signal and the vertically polarized complex baseband signal respectively to obtain the horizontally polarized power spectral density and the vertically polarized power spectral density; determining the horizontally polarized noise power and the vertically polarized noise power based on the horizontally polarized power spectral density and the vertically polarized power spectral density, and then calculating the estimated values ​​of the horizontally polarized noise power and the vertically polarized noise power for that time period; using the horizontally polarized power spectral density and the vertically polarized power spectral density, calculating the average echo power of the horizontally polarized and vertically polarized signals at each range gate respectively; combining the estimated values ​​of the horizontally polarized noise power and the vertically polarized noise power to obtain the horizontally polarized echo signal-to-noise ratio and the vertically polarized echo signal-to-noise ratio; and calculating the radar reflection based on the average echo power, the horizontally polarized echo signal-to-noise ratio, and the vertically polarized echo signal-to-noise ratio at each range gate. The radar reflectivity factor and differential reflectivity are calculated, and the average Doppler velocity and velocity spectral width are estimated based on the horizontal and vertical polarization power spectral densities. Odd-even pulse high-pass filtering is applied to the horizontal and vertical polarization complex baseband signals to remove ground clutter interference, resulting in filtered odd-even pulses. Cross-correlation and autocorrelation estimates are performed on the filtered odd-even pulses to obtain their respective cross-correlation and autocorrelation results. The cross-correlation results of the odd-even pulses are normalized using their corresponding autocorrelation results, and the absolute values ​​are averaged to obtain the correlation coefficient. The argument of the cross-correlation results of the odd-even pulses is calculated and averaged to obtain the initial differential propagation phase shift. The initial differential propagation phase shift is subjected to continuity processing to obtain the corrected differential propagation phase shift. Moving average smoothing is applied to the radar reflectivity factor, differential reflectivity, average Doppler velocity, and velocity spectral width. Moving median filtering smoothing is applied to the correlation coefficient and the corrected differential propagation phase shift to finally obtain the estimation results for all radar detection variables.

[0045] In some embodiments, the odd-even pulse high-pass filtering is implemented in one of the following two ways: DC removal: The average signal values ​​of the odd pulse group and the even pulse group in the horizontally polarized and vertically polarized complex baseband signals are calculated respectively. The average value is taken as the DC component of the corresponding pulse group. Then, the DC component of the pulse group to which each signal sample belongs is subtracted, and the AC component is retained to remove the main component of ground clutter; IIR high-pass filtering: A 6th-order elliptic filter is used. The stopband speed of the filter is set to 0.7 m / s and the passband speed is set to 1 m / s. A difference equation is constructed through preset autoregressive coefficients and moving average coefficients to filter the odd and even pulse groups respectively to completely suppress ground clutter.

[0046] In some embodiments, the specific process for determining the horizontal polarization noise power and the vertical polarization noise power is as follows: the horizontal polarization power spectral density and the vertical polarization power spectral density of each distance gate are sorted in ascending order to obtain the corresponding power spectral density sequence; the cumulative power sum of each sequential point in each sorted power spectral density sequence is calculated to form a power sum distribution map; based on the empirical rule that the noise power distribution is in the range of 5% to 40% power sum, the starting point where the power jump in the power sum distribution map exceeds a preset threshold is identified, and the cumulative power sum corresponding to the starting point is the horizontal polarization noise power or the vertical polarization noise power of the corresponding distance gate; the average horizontal polarization noise power and the average vertical polarization noise power of multiple distance gates and different radial directions are statistically analyzed and used as the estimated values ​​of the horizontal polarization noise power and the vertical polarization noise power for that time period, respectively.

[0047] In some embodiments, after performing discrete Fourier transform on the horizontally polarized complex baseband signal and the vertically polarized complex baseband signal, a ground clutter suppression step is further included: determining the digital frequency half-bandwidth corresponding to the ground clutter, and the power spectral density of the target echo signal adjacent to the low-frequency end and high-frequency end of the ground clutter; calculating the rate of change of the signal power spectral density based on the target echo signal power spectral density; and using the linear interpolation result corresponding to the rate of change to replace the value of the part of the power spectral density interfered with by the ground clutter, to obtain the horizontally polarized power spectral density and the vertically polarized power spectral density after clutter removal.

[0048] In some embodiments, the cross-correlation estimation is performed separately according to the grouping method of odd pulse short interval and even pulse long interval; the autocorrelation estimation is performed separately according to the polarization type of horizontal polarization and vertical polarization, combined with the grouping method of odd pulse short interval and even pulse long interval, and the autocorrelation estimation result is used to characterize the echo power of the corresponding pulse group; the absolute value is taken after normalization processing to limit the correlation coefficient to the real number range of 0 to 1, so as to avoid the small imaginary components generated by complex number calculation from affecting the estimation result.

[0049] In some embodiments, the continuous processing of the initial differential propagation phase shift specifically includes: eliminating transient jumps: detecting the jump amplitude of the initial differential propagation phase shift between adjacent range gates; if the jump amplitude exceeds the phase jump perturbation threshold but is less than the maximum phase jump range threshold, then correcting the initial differential propagation phase shift of the current range gate to the value of the previous range gate; eliminating out-of-range jumps: if the jump amplitude exceeds the phase forward folding threshold or the phase reverse folding threshold, adjusting the initial differential propagation phase shift of the next range gate and recalculating the jump amplitude, replacing abnormal measurements with valid measurements from neighboring range gates; and normalizing the numerical range: if the initial differential propagation phase shift is greater than 180°, subtracting 360°; if it is less than -180°, adding 360° to ensure that the corrected differential propagation phase shift is within the range of ±180°.

[0050] In some embodiments, the number of sliding averages in the sliding average smoothing process is set to 10 distance gates; the processing length of the sliding median filter smoothing process corresponds to the number of distance gates in the distance range of 300m to 500m, so as to eliminate the small fluctuations generated during the processing and ensure that the estimation results are smooth and undistorted.

[0051] In some embodiments, the Discrete Fourier Transform adapts to two operating modes: single-pulse repetition interval and staggered-pulse repetition interval. When the radar operates in single-pulse repetition interval mode, Fast Fourier Transform is used to process the horizontally polarized complex baseband signal and the vertically polarized complex baseband signal to directly obtain the horizontally polarized power spectral density and the vertically polarized power spectral density. When the radar operates in staggered-pulse repetition interval mode, staggered Discrete Fourier Transform is used to process odd-pulse groups and even-pulse groups separately to ensure the uniformity of the sampling interval, thereby obtaining the horizontally polarized power spectral density and the vertically polarized power spectral density.

[0052] In some embodiments, when using staggered discrete Fourier transform technology, the image frequency component removal step is further included: for each distance gate, the position of the strongest power spectral density among the horizontal polarization power spectral density and the vertical polarization power spectral density is found; based on the characteristic that the amplitude of the main frequency region is greater than the amplitude of the image frequency component, the effective frequency domain range is determined with the position of the strongest power spectral density as a reference, and the image frequency component generated by zero-padding interpolation is removed to obtain a pure power spectral density.

[0053] In some embodiments, the calculation of the differential reflectivity further includes attenuation correction, channel error correction, and observation error correction: attenuation correction is performed based on the attenuation coefficient determined by fitting after actual observation; channel error correction describes the power measurement difference between horizontal and vertical polarization within the normal linear dynamic range of the radar using a quadratic fitting curve determined during factory calibration; observation error correction is performed based on the observation error value determined by field calibration to improve the estimation accuracy of differential reflectivity.

[0054] The technical concept of this invention is as follows:

[0055] The design adopts the approach of "time-frequency domain divide-and-conquer + reuse of previous technologies", and the specific solution is as follows:

[0056] Technical basis: Obtaining high sampling rate power spectral density information to support subsequent processing;

[0057] Frequency domain processing: Based on power spectral density analysis, two types of variables are estimated: one is amplitude characteristic-related variables (horizontal / vertical polarization echo signal-to-noise ratio, reflectivity factor, differential reflectivity), and the other is velocity-related variables (mean Doppler velocity, velocity spectral width). Finally, smoothing is performed using moving average.

[0058] Time-domain processing: The horizontal / vertical polarized complex baseband signal is grouped into odd and even pulses, and ground clutter is suppressed by DC removal or IIR high-pass filtering. The correlation coefficient and differential propagation phase shift are obtained by cross-correlation / autocorrelation estimation and then optimized by continuity correction (eliminating jumps and normalizing the range) and moving median filtering.

[0059] Adaptive design: Fast Fourier Transform is used in single-pulse repetition interval mode, and staggered Discrete Fourier Transform is used in staggered pulse repetition interval mode, with the rest of the processing logic remaining the same.

[0060] The first paper estimates radar detection variables with pure amplitude characteristics;

[0061] Step 1: Perform Discrete Fourier Transform on the horizontally and vertically polarized radar complex baseband signals respectively;

[0062] Performing Discrete Fourier Transform on the horizontally polarized and vertically polarized complex baseband signals output from the radar digital receiver yields the following results: Figure 2 and Figure 3 The measured results of the horizontal and vertical polarization power spectral densities under the staggered pulse repetition intervals shown are denoted as follows: and .

[0063] Step 2: Determine the noise power of the horizontally polarized and vertically polarized echoes in the power spectral density;

[0064] from Figure 4 As can be seen, signals below a certain power spectral density remain almost unchanged. The sum of the powers below this threshold is the noise power in the frequency domain.

[0065] First, the power spectral density at a certain distance gate r = R is sorted from low to high to obtain a new power spectral density sequence. N is the total number of frequency points in the frequency domain; then, the power sum of each sequence point is calculated, i.e. , can be obtained as Figure 5 The power distribution diagram is shown. Clearly, the power sum reaches 100% at point N. Based on empirical statistics, noise power generally falls within the 5% to 40% power sum range, and the transition from noise power to signal power is a jump, denoted as a transition, which is not less than... δ N For weather radar, δ N Approximately 1 dB.

[0066] Through comparison and judgment, power and jump exceed δ N From the beginning of the transition, all signals are either signals or interference; and the starting point of the transition, that power sum, is the noise power measured in the frequency domain. P N For a dual-polarization radar, the noise powers of its horizontal and vertical polarizations are respectively... P HN and P VN .

[0067] Generally, within a certain time period, the noise power measured in the frequency domain for different range gates is almost the same. Therefore, the average noise power of several range gates can be used as the noise power of that radial scan. Similarly, within a certain time period, the noise power differences across different radial directions are also small, so further averaging can be performed to obtain an estimate of the noise power for that time period. P NE .

[0068] In this way, noise power estimation can be used even in the absence of a signal. P NE This serves as the basis for determining the presence or absence of echo signals. For dual-polarization radar, the horizontal and vertical polarization characteristics are as follows: P HNE and P VNE .

[0069] Step 3: Determine the horizontal and vertical polarization echo power of the precipitation target in the power spectral density.

[0070] from Figure 4 The given power spectral density shows that the average echo power at a certain distance gate r is:

[0071] ;

[0072] This power includes noise power. Since the power in the time domain is equal to the power in the frequency domain, this power can serve as a direct data source for estimating radar detection variables. For a dual-polarization radar, the average echo power of its horizontal and vertical polarizations at a certain range gate r are respectively:

[0073] ;

[0074] As an example, the average echo power distribution of horizontal polarization at a certain distance gate r is shown below. Figure 6 The upper part.

[0075] Step 4: Calculate the radar reflectivity factor using the horizontally polarized echo power;

[0076] From the average echo power of the horizontal and vertical polarization radars at a certain range gate r, their signal-to-noise ratios can be calculated as follows:

[0077] ;

[0078] In the absence of radar echoes, i.e., when the result of the numerator on the right side of the above equation is not greater than 0, or when the noise power in a separate radial direction cannot be obtained, its (horizontally polarized or vertically polarized) signal-to-noise ratio is recorded as 0 dB (the average echo power is equal to twice the noise power), and its noise is recorded as... P HNE or P VNE (This is the linear value without taking the logarithm).

[0079] According to the weather radar equation, taking the logarithm and simplifying, we get:

[0080] ;

[0081] in, Zh(r) The distribution of the horizontal polarization reflectivity factor with distance r (in km) in a certain radial direction is given by the expression. dBZ The first term on the right side of the equation, SNR H (r) The distribution of horizontally polarized signal-to-noise ratio with distance, in units of dB ; Item 2, dBZ 0 represents the radar performance characterization value, in units of... dBZ Specifically, it is expressed as:

[0082] ;

[0083] in, These are the factory calibration test values ​​for the radar, characterizing the radar measurement sensitivity at a reference distance (1 km) during calibration. (Unit: ...) dBZ ;and radar constant CdBZ The calculation method is as follows:

[0084] ;

[0085] In the above formula, the variables and their dimensions are as follows: λ is the radar operating wavelength, in cm; P t τ represents the radar's peak transmitted power, in kW; τ represents the radar's transmitted pulse width, in µS; θ and φ represent the radar antenna's azimuth and elevation beamwidths, respectively, in µS. ; L T G represents the total transmit and receive attenuation, which is a linear value > 1. T G R The transmit and receive gains of the radar antenna are expressed linearly.

[0086] and I dB C The radar system sensitivity is calculated to the feed point, in units of... dBm This was obtained through calibration testing; N dB r The noise power measured during the observation, in units of dBm ; N dB C The noise power is measured during factory calibration, in units of... dBm .

[0087] Radar constant dBZ The larger the value of 0, the worse the radar performance.

[0088] Item 3, Z offset Total radar deviation, in units of dBZ Including systematic bias Z 0s (Caused by the change in system gain over time, obtained from instrument testing) and observation bias Z 0o (Obtained from field calibration); default values ​​are all 0. Item 4, RC For distance correction, for volume targets, the correction value is the base-10 logarithm of the square of the target distance in km. (Item 5) CCOR Ground clutter correction factor, in units dB Let the powers of suppressed and unsuppressed ground clutter be respectively... R 0 and T 0, then Item 6, A ( r Corrections for rain area attenuation. ,α This is the attenuation coefficient, which needs to be determined through fitting after actual observation. For the X-band, it can be approximately taken as 0.14 to 0.35 dB / °, with an average of 0.28 dB / °. φ DP C For differential propagation phase shift, it is derived from the differential dissimilarity rate. K DP Obtained through integration.

[0089] Similarly, the distribution of the radial vertical polarization reflectivity factor with distance r (unit: km) Zv(r) It can also be estimated. However, this value is not used in many applications.

[0090] Step 5: Calculate the differential reflectivity using the echo power of horizontal and vertical polarization;

[0091] By definition, the method for estimating the measured value of differential reflectance is as follows:

[0092] ;

[0093] As an example, the distribution of differential reflectance measurements estimated at a certain distance gate r is shown below. Figure 6 The lower part.

[0094] In practical applications, system bias corrections and compensation for the difference in horizontal and vertical polarization attenuation caused by rainfall must be considered. Therefore, the differential reflectance estimation method needs to be modified.

[0095] ;

[0096] The second item on the right is the attenuation correction. β This is the attenuation coefficient, which needs to be determined by fitting after actual observation. For the X-band, it can be approximately 0.03 to 0.06 dB / °, with an average of 0.05 dB / °. φ DP C As mentioned above, this is differential propagation phase shift, from K DP The result is obtained by integration, in degrees (°).

[0097] The third item on the right is channel correction. δ ch The channel error, in dB, can be described by a quadratic fit to account for the power measurement differences between horizontal and vertical polarizations within the normal linear dynamic range of the radar. The quadratic curve for channel calibration has been determined as {a, b, c} during factory calibration.

[0098] Item 4 on the right δ Zdr OThis value, in dBZ, can be determined by field calibration to correct for observation errors.

[0099] The second part estimates radar detection variables related to velocity;

[0100] Since the power spectral density distance distribution with horizontal polarization has been obtained as described above, it becomes relatively easy to estimate the average Doppler velocity and velocity spectral width in the frequency domain.

[0101] Step 6: Determine the mean Doppler velocity of the precipitation target in the power spectral density;

[0102] The formula for estimating the average Doppler velocity based on the definition of power-weighted average is as follows:

[0103] ;

[0104] Thus, the distribution of the average Doppler velocity with distance was obtained. See the practical example below. Figure 8 The portion of the curve with the largest variation represents the average Doppler velocity.

[0105] Step 7: Determine the velocity spectral width of the precipitation target in the power spectral density;

[0106] The formula for estimating the root mean square error of velocity based on the definition of power-weighted average is as follows:

[0107] ;

[0108] The formula for converting the velocity mean square error into velocity spectral width is as follows:

[0109] ;

[0110] See actual examples Figure 8 The portion of the curve with smaller changes represents the velocity spectral width estimate.

[0111] Step 8: Data smoothing.

[0112] Especially for short-wavelength radars like those in the X-band, where the power spectral density exhibits some fluctuation with distance, to ensure smooth display and observation, the reflectivity factor—the average echo power distribution—is used as a pre-estimated variable in the four basic data measurements of the radar described earlier. and The pre-estimation variable for differential reflectance is the measurement of differential reflectance. Mean Doppler velocity and velocity spectrum A moving average that varies with distance needs to be performed. The method is to perform a moving average smoothing on these variables with the number of moving averages N_smooth (which is also the number of distance gates, usually 10), which can yield a smoother data curve without distortion.

[0113] Part Three: Estimation of Radar Detection Variables with Almost No Impact on Amplitude;

[0114] Theoretical analysis shows that, under a certain signal-to-noise ratio (SNR), the cross-correlation estimation is not directly related to the absolute amplitude of the signal. Therefore, if the signal is filtered, its intensity will be reduced (amplitude decreases), but its SNR will still far exceed the required minimum threshold. This has no impact on the cross-correlation estimation, but it eliminates the negative impact of interference clutter on the processing, making the estimation more realistic.

[0115] Step 9: Perform odd-even pulse high-pass filtering on the horizontally and vertically polarized radar complex baseband signals respectively;

[0116] In this step, two methods are used to implement high-pass filtering: DC removal and IIR high-pass filtering. The former may leave a small amount of ground clutter components that do not affect the estimation results, but the processing is extremely simple and has no delay effect. The latter removes the substrate more thoroughly, but the processing is more complex, and the delay problem must be solved. In addition, the nonlinear phase of IIR filters has a slight impact on signal quality. The choice depends on the different radar applications. For example, the former is more suitable for weather radars with high requirements for amplitude and phase.

[0117] Because this part of the processing does not involve calculating amplitude-related variables, slight changes in the absolute amplitude of the signal will not produce visible bias in the variable estimation results.

[0118] DC removal is straightforward. Based on the principle that the average of a sine wave—the most typical AC signal—is zero, the average result is the DC component of the signal. Subtracting the DC component from each sample leaves only the AC component. The DC component is the main component of ground clutter. This method easily removes the influence of ground clutter.

[0119] It's important to note that for single-pulse repetition intervals, simply averaging is sufficient. However, for staggered pulse repetition intervals, the odd and even pulse cases must be handled separately to ensure the uniformity of the sampling intervals, thus validating the principle that the mean is DC.

[0120] The IIR high-pass filter scheme first designs a ground clutter IIR high-pass filter with a stopband velocity of Vrs (which can be set to 0.7 m / s) and a passband velocity of Vrp (which can be set to 1 m / s). Following the design method of efficient elliptic filters, the autoregressive coefficients of the 6th-order filter are [1.0000 -3.8364 6.5833 -6.3152 3.5951 -1.1659 0.1832086], and the moving average coefficients are [0.3534921 -2.0440 4.9991 -6.6171 4.9991 -2.0440 0.3534921].

[0121] To speed up the process, the sampling method in the filter settings is used to trade space for time. This requires more storage units but significantly reduces computation time.

[0122] The difference equation implemented by the filter is as follows:

[0123] ;

[0124] IIR filter signal flow for suppressing ground clutter, such as Figure 9 Its frequency domain amplitude response characteristics are shown in Figure 10 .

[0125] It should be noted that if blind speed occurs during radar detection, although the intensity will be reduced, this filtering does not affect the accuracy of the results, just like the occurrence of real ground clutter.

[0126] Step 10: Estimate the cross-correlation and autocorrelation of the odd and even pulses respectively;

[0127] After the dual-polarized radar echo signal is processed by ground clutter filtering, cross-correlation and estimation of their respective autocorrelation parameters can be performed, which are not significantly related to power.

[0128] The zero-order cross-correlation estimate is performed using odd (short interval) and even (long interval) pulses respectively, as shown in the following equation:

[0129] ;

[0130] Autocorrelation is essentially the calculation of echo power, performed for horizontal and vertical polarization, and for odd (short interval) and even (long interval) pulses, as shown in the following formula:

[0131] ;

[0132] Step 11: Combine the odd-even cross-correlation estimation results and convert them into correlation coefficients and differential propagation phase shifts;

[0133] After the aforementioned steps, the odd-even cross-correlation estimate and its echo power result are obtained. According to the definition of the correlation coefficient of dual-polarized weather radar, its zero-order cross-correlation is normalized by the echo power of the odd and even pulses respectively. Considering that the correlation coefficient is theoretically a real number in the range of [0, 1], but complex calculation may produce a very small imaginary number, the absolute value of the processing result is taken.

[0134] ;

[0135] Then, the average is calculated, which completes the estimation of the correlation coefficient.

[0136] ;

[0137] After this processing, the correlation coefficient will not show large abrupt changes, but small variations similar to noise will still exist. After another moving median filter, the correlation coefficient will become smoother and closer to the truth.

[0138] Similarly, based on the definition of differential propagation phase shift in dual-polarized weather radar, the zero-order cross-correlation is used to calculate the argument angles, and then the average of the odd and even pulse arguments is calculated. This is then converted into a unit of degrees (°) to obtain an estimate of the differential propagation phase shift.

[0139] ;

[0140] Step 13: Determine the correct estimate of the differential propagation phase shift;

[0141] Because the signal-to-noise ratio in the echo signal may be extremely low (e.g., less than 10 dB required by dual-polarized radar), or adverse conditions such as co-channel or adjacent-channel interference may occur, the cross-correlation estimation will fluctuate or even be incorrect. As a result, the differential propagation phase shift will also fluctuate or even be incorrect. However, in actual signals, if this situation does not occur, the differential propagation phase shift exhibits spatiotemporal continuity.

[0142] Therefore, it can be seen that performing continuous processing on the phase shift of differential propagation can make it more realistic.

[0143] The continuous processing of differential propagation phase shift includes three aspects: eliminating transient jumps, eliminating out-of-range jumps, and normalizing the numerical range.

[0144] Eliminating transient jumps is based on the theoretical basis and fact that the differential propagation phase shift will not exceed a certain threshold in the distance distribution. The degree of jump in distance is detected, while also considering phase ambiguity (which causes folding beyond 360°). The processing flow is as follows: Figure 11 As shown.

[0145] Eliminating out-of-range jumps is also based on the theoretical basis and facts of continuity. It involves detecting the degree of jump in distance; if the jump exceeds a threshold, it may be caused by factors such as extremely low signal-to-noise ratio or the presence of co-channel or adjacent-channel interference in the signal. In such cases, the measured value may not represent the true situation, and using a nearby measurement is the best choice. In this situation, the jump must also be processed according to the actual situation. See the detailed process below. Figure 12 The flowchart shown is shown.

[0146] The numerical range normalization process is much simpler, because the measured phase shift value of differential propagation will not exceed [the specified value]. This is considered within a 180° range. Based on the first two adjustments, the differential propagation phase shift may exceed... The value range is 180°. Therefore, this problem can be easily handled according to the phase folding principle; see details. Figure 13 The flowchart shown is shown.

[0147] After this processing, the differential propagation phase shift will not show large abrupt changes, but small variations similar to noise will still exist. After another sliding median filter, the differential propagation phase shift will become smoother and closer to reality.

[0148] This yields accurate estimates of the correlation coefficient and differential propagation phase shift of the dual-polarized radar, independent of ground clutter interference.

[0149] Example 1: Estimation of detection variables for a dual-polarization Doppler weather radar operating in single or staggered pulse repetition interval mode;

[0150] First, the power spectral density range distribution of a dual-polarization Doppler weather radar operating in a staggered pulse repetition interval mode with a stagger ratio of n1:(n1+1) is estimated. This forms the basis for estimating the reflectivity factor, mean Doppler velocity, velocity spectral width, and differential reflectivity detection variables in the frequency domain.

[0151] In the time-domain processing, the amplitude-sensitive variables mentioned above are avoided. Only the 0th-order cross-correlation is used to form a correlation coefficient probe variable that has its power influence eliminated through normalization, and a phase parameter—the differential propagation phase shift probe variable—that is independent of amplitude under certain signal-to-noise ratio conditions. Since high-quality processing methods already exist for differential propagation phase shift rates, they will not be elaborated upon here.

[0152] In this embodiment, the most commonly used staggered pulse repetition interval, n1=4, n1:(n1+1)=4:5, is used to estimate the detection variables of the dual-polarized Doppler weather radar, and the corresponding process is demonstrated. For the case of a uniform pulse repetition interval, the Fourier transform is simpler and can be directly implemented using the Fast Fourier Transform. The remaining processing is similar and will not be specifically listed here.

[0153] The horizontally polarized and vertically polarized complex baseband signals output from the radar signal processor, according to Figure 1 The method involves processing the data in two parts: the frequency domain and the time domain.

[0154] Step 1: Perform Discrete Fourier Transform on the horizontally and vertically polarized radar complex baseband signals respectively;

[0155] Radar echo data is a complex signal with horizontal / vertical polarization input. x H / x V ,Right now:

[0156] ;

[0157] in, n To handle the pulse sequence within the resolution angle, n p = 1，2，…，N P , N P To process the total number of pulses within the resolution angle; n r To handle the distance gate (library) order, n r = r / δr, r For distance, δr This refers to distance resolution. The same applies below.

[0158] Estimation of horizontal / vertical polarized two-dimensional power spectral density Sp H / Sp V If it is a single-pulse repetition interval, the amplitude value of the sequential pulse Hanning window weighted discrete Fourier transform (DFT) of the horizontal / vertical polarized complex signal with the same distance gate (ku) is:

[0159] ;

[0160] in, w h For Hanning window, W It is a disc-shaped factor.

[0161] If the radar uses staggered pulse repetition intervals, the conventional Fourier transform is unusable because the sampling interval is not constant. In this case, the staggered discrete Fourier transform technique must be used, resulting in a frequency domain signal that contains only amplitude and not phase. The staggered discrete Fourier amplitude transform (SDFT) yields the horizontally polarized power spectral density. With horizontal polarization noise power P HNE and vertical polarization power spectral density With vertical polarization noise powerP VNE .

[0162] ;

[0163] In the formula, the superscript on the left indicates the staggered pulse repetition interval, the subscript indicates polarization, and the left side represents the power spectral density at the distance gate. n r Distribution above, k These are discrete frequencies. (Right side) SDFT The symbol for the discrete amplitude Fourier transform of the staggered pattern is... n 1 and n 2 represents the ratio of inconsistencies, and n 2 = 1 + n 1. Let . n p = 0, 1, 2, ..., N p -1 , N p It must be an even number. The implementation method of staggered discrete amplitude Fourier transform (SDFT) is as follows:

[0164] First, perform zero-padding interpolation on the signal and then perform a windowed Fourier transform to obtain the discrete Fourier transform V of the zero-padding sequence. Taking a 4:5 pulse repetition interval and horizontal polarization as an example... n 1 = 3 n 2 = 4. The signal is expressed as:

[0165] ;

[0166] Among them, the superscript on the right T Indicates transpose; This indicates zero-padding interpolation, representing the discrete-time arrangement of the rearranged signal; This indicates a rearranged sequence. The length of the interpolated sequence becomes... .

[0167] A weighted discrete Fourier transform is performed on this new sequence. The weighting is done using von Neumann. von Hann window, that is:

[0168] ;

[0169] The frequency domain column vector is then generated as follows:

[0170] ;

[0171] Second, the column vector V H Decomposed into The matrix, For the number of pulse logarithms, that is, rearranged as :

[0172] ;

[0173] in, V i column vector V h Decomposed into m OK j Column elements, m = 0, 1, …, p -1, j = 0, 1, …, The position in the column vector is .

[0174] Third, use constants p p Reverse Formation matrix Convert to a staggered discrete amplitude Fourier transform matrix.

[0175] ;

[0176] Fourth, convert the staggered discrete Fourier amplitude matrix into a staggered discrete amplitude Fourier transform column vector. Although the matrices are different, methodologically, it is the reverse of the second step. Expanded representation:

[0177] ;

[0178] Clearly, rearranging the elements of this matrix transforms its row and column positions into column vector positions. This yields the staggered discrete amplitude Fourier transform column vector. Its elements are:

[0179] ;

[0180] Fifth, remove the image frequency components of the discrete amplitude Fourier transform caused by regular zero-padding interpolation.

[0181] The aforementioned processing illustrates the staggered discrete amplitude Fourier transform result of a certain distance gate. If the distance distribution is taken into account, it can be expressed as:

[0182] Considering that the amplitude of the main frequency region is always greater than the amplitude of the image frequency component, and that the width of each frequency domain is 1 / 3 of the total frequency domain width. p The actual situation can be handled in this way to obtain the result of the discrete amplitude Fourier transform of the staggered amplitude.

[0183] First, find the location of the strongest power spectral density at a certain distance gate:

[0184] ;in, This is the maximum value position operator.

[0185] Then, determine the results of the staggered discrete amplitude Fourier transform:

[0186] ;in, .

[0187] The vertical polarization processing method is similar.

[0188] Step 2: Calculate the horizontal / vertical polarization clutter power spectral density. ;

[0189] Let the digital frequency half-bandwidth of the ground clutter be... Its power spectral density at the low-frequency end is Then move one more point to the left. The power spectral density is the power spectral density of the target echo signal. The power spectral density at the high-frequency end is Then move one more point to the right. The power spectral density is the power spectral density of the target echo signal. The rate of change of the signal power spectral density can be obtained from the power spectral density of the echo signals at both ends and the digital frequency bandwidth of the ground clutter. .

[0190] ;

[0191] In this way, by using the power spectral density obtained by processing with this rate of change to replace the part with ground object interference, the ground clutter interference suppression is completed.

[0192] The vertical polarization processing method is similar.

[0193] Horizontal and vertical polarization clutter power spectral densities are as follows Figure 2 and 3 As shown.

[0194] Step 3: Estimate the horizontal / vertical polarization clutter interference signal-to-noise ratio ;

[0195] Figure 4 The power spectral density at a range gate in the frequency domain, as well as the characterization of the average Doppler velocity and velocity spectral width, are clearly presented. It is evident that the horizontal / vertical polarized clutter echo power can be calculated from the clutter power spectral density.

[0196] ;

[0197] Figure 5 This provides a schematic diagram of the power spectral density sorted from low to high at a distance gate in the frequency domain, as well as the characteristics that distinguish the signal from noise. In the frequency domain, power is calculated from the power spectral density by direct summation. Therefore, the noise power level can be easily estimated. Theoretically, the noise power is located at approximately 10% of the lower end of the power spectral density.

[0198] ;

[0199] here, N f This refers to the total number of digital frequency points in the frequency domain. In the single-pulse repetition interval Fourier transform, this value is... N p In the Fourier transform of the amplitude of the staggered pulse repetition interval, this value is... N p / 2, and its processing range is selected from the main frequency range.

[0200] Finally, the signal-to-noise ratio is estimated, which can be done using either a linear value for calculation or a logarithmic value for expression.

[0201] ;

[0202] in, SNR th The identification coefficient can be 1 to 3 dB, but 3 dB is commonly used.

[0203] The vertical polarization processing method is similar. Therefore, we can obtain... Figure 6 Signal-to-noise ratio estimation results for the frequency domain processing of the significantly varying portion.

[0204] Step 4: Reduce clutter signal-to-noise ratio by horizontal polarization. SNR h Calculate the radar reflectivity factor;

[0205] This is quite simple; the calculation can be done directly using the aforementioned formula. It should be noted that the precipitation echo in this example is weak, so its impact on echo attenuation is difficult to observe, and the compensation part is almost entirely displayed. Reflectivity factor Z h The results are as follows Figure 7 As shown.

[0206] Step 5: Calculate the differential reflectivity using the echo power of horizontal and vertical polarization;

[0207] First, calculate the differential signal-to-noise ratio for horizontal / vertical polarization clutter removal. SNR dr Then, the differential reflectance is calculated.

[0208] ;

[0209] If the noise levels of horizontal / vertical polarization are the same, approximately equal to becomes an equal sign or equal to. Differential reflectivity, such as Figure 6 The flat section is shown.

[0210] Step 6: Calculate the average Doppler velocity of precipitation using the power spectral density of horizontally polarized clutter removal;

[0211] The radial velocity of the target or its radial velocity component is the average Doppler velocity, which can be calculated using the following formula.

[0212] ;

[0213] in, The distribution of Doppler velocity in the velocity dimension measured by radar. The minimum speed (the negative maximum unambiguous speed). The maximum speed (positive maximum unambiguous speed). Speed ​​is 0.

[0214] The mean Doppler velocity calculation results are as follows: Figure 8 The parts that have changed the most are shown.

[0215] Step 7: Calculate the velocity spectrum width of precipitation using the power spectral density of horizontally polarized clutter removal;

[0216] The method for calculating the velocity spectral width is as follows:

[0217] ; For velocity variance, spectral width W s The standard deviation is denoted as .

[0218] The calculation results of the velocity spectral width are as follows: Figure 8 The parts with minor changes are shown.

[0219] Step 8: Calculate the 0th order cross-correlation coefficient of horizontal / vertical polarization to DC using the time-domain method. ;

[0220] Theoretical analysis shows that, under a certain signal-to-noise ratio (SNR), the cross-correlation estimation is not directly related to the absolute amplitude of the signal. Therefore, if the signal is filtered, its intensity will be reduced (amplitude decreases), but its SNR will still far exceed the required minimum threshold. This has no impact on the cross-correlation estimation, but it eliminates the negative impact of interference clutter on the processing, making the estimation more realistic.

[0221] Estimating the 0th-order cross-correlation (sometimes simply called the correlation coefficient) of a DC signal will result in a large deviation. However, due to the alternating positive and negative characteristics of AC signals, they will self-cancel out over a longer period (multiple pulses), thus having little impact. The method for estimating the DC component of an input horizontal / vertical polarized complex signal is as follows:

[0222] ;

[0223] This removes the DC component of the horizontal / vertical polarized complex signal, resulting in an AC signal.

[0224] ;

[0225] At this point, cross-correlation can be calculated: ;

[0226] Finally, calculate the cross-correlation coefficient, which is the absolute value of the cross-correlation normalized to their respective powers:

[0227] ;

[0228] It's important to note that the processing methods differ slightly between the single-pulse repetition interval and the staggered-pulse repetition interval working modes. The former can be calculated directly using the formula; the latter requires processing by grouping odd and even pulses, with each group's length becoming half the total number of pulses, and the correlation coefficients of each group being averaged for the output result.

[0229] The calculation results of the autocorrelation coefficient are as follows: Figure 14 As shown. Note that the autocorrelation coefficient tends to 0 because there is no signal, the signal-to-noise ratio is not greater than 0, and it is due to noise. In practical applications, meaningless displays can be easily removed.

[0230] Step 9: Determine the correct estimate of the differential propagation phase shift;

[0231] The previous step provided a cross-correlation estimate, whose argument is the differential propagation phase shift. The calculated value for the single-pulse repetition interval can be used directly. For staggered pulse repetition intervals, the odd and even cross-differential propagation phase shift results need to be averaged and combined. After combination, according to the aforementioned method, the distance distribution of the differential propagation phase shift is processed for continuity, resulting in the following... Figure 15 The differential propagation phase shift results are shown. Note that the abrupt changes in the differential propagation phase shift occur because there is no signal, and the signal-to-noise ratio is not greater than 0, which is due to noise. In practical applications, meaningless displays can be easily removed.

[0232] Step 10, data smoothing;

[0233] To ensure the practicality of the estimated variables, it is necessary to eliminate minor fluctuations generated during processing. This requires smoothing using moving averages or moving median filtering. In this embodiment, the time-domain processing portion, i.e., the cross-correlation portion, is processed using moving median filtering; the frequency-domain processing portion, i.e., the remaining portion, is processed using moving averages. The processing length is set to a number of points between 300m and 500m. The results are as follows... Figure 6 , 8 The smooth sections shown in 14 and 15.

Claims

1. A method for estimating detection variables in a time-frequency domain dual-polarization radar with staggered pulse repetition intervals, characterized in that, include: The horizontally polarized complex baseband signal and the vertically polarized complex baseband signal output by the radar digital receiver are acquired. Discrete Fourier transforms are performed on the horizontally polarized complex baseband signal and the vertically polarized complex baseband signal respectively to obtain the horizontally polarized power spectral density and the vertically polarized power spectral density. Based on the horizontal polarization power spectral density and the vertical polarization power spectral density, the horizontal polarization noise power and the vertical polarization noise power are determined, and then the estimated values ​​of the horizontal polarization noise power and the estimated values ​​of the vertical polarization noise power are calculated. Using the horizontal polarization power spectral density and the vertical polarization power spectral density, the average echo power of horizontal polarization and vertical polarization at each distance gate is calculated respectively. The signal-to-noise ratio of horizontal polarization echo and the signal-to-noise ratio of vertical polarization echo are obtained by combining the estimated values ​​of horizontal polarization noise power and vertical polarization noise power. Based on the average echo power of horizontal and vertical polarization at each range gate, the signal-to-noise ratio of horizontal polarization echo, and the signal-to-noise ratio of vertical polarization echo, the radar reflectivity factor and differential reflectivity are calculated. At the same time, the average Doppler velocity and velocity spectral width are estimated based on the power spectral density of horizontal polarization and the power spectral density of vertical polarization. The horizontally polarized complex baseband signal and the vertically polarized complex baseband signal are respectively subjected to odd-even pulse high-pass filtering to remove ground clutter interference and obtain the filtered odd-even pulse; Cross-correlation and autocorrelation estimates are performed on the filtered odd and even pulses respectively to obtain the cross-correlation and autocorrelation results for the odd and even pulses. The cross-correlation results of the odd and even pulses are normalized with the corresponding autocorrelation results, and the absolute values ​​are taken and averaged to obtain the correlation coefficient. The argument of the odd-even pulse cross-correlation results is calculated and averaged to obtain the initial differential propagation phase shift; The initial differential propagation phase shift is processed to obtain the corrected differential propagation phase shift; The radar reflectivity factor, differential reflectivity, average Doppler velocity, and velocity spectrum width are smoothed by moving average. The correlation coefficient and the corrected differential propagation phase shift are smoothed by moving median filtering. Finally, the estimation results of all radar detection variables are obtained.

2. The method for estimating detection variables in the time-frequency domain of a dual-polarization radar with staggered pulse repetition intervals according to claim 1, characterized in that, The odd-even pulse high-pass filter is implemented using one of the following two methods: DC removal processing: Calculate the average signal value of the odd pulse group and even pulse group in the horizontally polarized complex baseband signal and the vertically polarized complex baseband signal respectively. Take the average value as the DC component of the corresponding pulse group. Then, subtract the DC component of the pulse group to which each signal sample belongs, and retain the AC component to remove the main component of ground clutter. IIR high-pass filtering: A 6th-order elliptic filter is used. The stopband speed of the filter is set to 0.7 m / s and the passband speed is set to 1 m / s. A difference equation is constructed by using preset autoregressive coefficients and moving average coefficients to filter the odd and even pulse groups respectively to completely suppress ground clutter.

3. The method for estimating the detection variables of a time-frequency domain dual-polarization radar with staggered pulse repetition intervals according to claim 1, characterized in that, The specific process for determining the horizontal polarization noise power and the vertical polarization noise power is as follows: The horizontal polarization power spectral density and vertical polarization power spectral density of each distance gate are sorted in ascending order to obtain the corresponding power spectral density sequence. Calculate the cumulative power sum of each sequential point in each sorted power spectral density sequence to form a power sum distribution map; Based on the empirical rule that noise power distribution ranges from 5% to 40% of the power range, the starting point where the power jump in the power distribution map exceeds the preset threshold is identified. The cumulative power at this starting point is the horizontal polarization noise power or vertical polarization noise power of the corresponding distance gate. The mean horizontal polarization noise power and the mean vertical polarization noise power of multiple distance gates and different radial directions are statistically analyzed and used as the estimated values ​​of horizontal polarization noise power and vertical polarization noise power, respectively.

4. The method for estimating the detection variables of a time-frequency domain dual-polarization radar with staggered pulse repetition intervals according to claim 1, characterized in that, After performing discrete Fourier transform on the horizontally polarized complex baseband signal and the vertically polarized complex baseband signal, the process also includes a ground clutter suppression step: Determine the digital frequency half-bandwidth corresponding to the ground clutter, and the power spectral density of the target echo signal adjacent to the low-frequency end and high-frequency end of the ground clutter; Calculate the rate of change of the signal power spectral density based on the target echo signal power spectral density; The linear interpolation result corresponding to this rate of change is used to replace the value of the power spectral density affected by ground clutter, thus obtaining the horizontal polarization power spectral density and vertical polarization power spectral density after clutter removal.

5. The method for estimating the detection variables of a time-frequency domain dual-polarization radar with staggered pulse repetition intervals according to claim 1, characterized in that, The cross-correlation estimation is performed separately according to the grouping method of odd pulse short interval and even pulse long interval; the autocorrelation estimation is performed separately according to the polarization type of horizontal polarization and vertical polarization, combined with the grouping method of odd pulse short interval and even pulse long interval, and the autocorrelation estimation results are used to characterize the echo power of the corresponding pulse group. The absolute value is taken after normalization to limit the correlation coefficient to the real number range of 0 to 1, so as to avoid the small imaginary components generated by complex number calculation from affecting the estimation results.

6. The method for estimating the detection variables of a time-frequency domain dual-polarization radar with staggered pulse repetition intervals according to claim 1, characterized in that, The continuous processing of the initial differential propagation phase shift specifically includes: Eliminate transient jumps: Detect the jump amplitude of the initial differential propagation phase shift between adjacent range gates. If the jump amplitude exceeds the phase jump perturbation threshold but is less than the maximum phase jump range threshold, then correct the initial differential propagation phase shift of the current range gate to the value of the previous range gate. Eliminating out-of-range jumps: If the jump amplitude exceeds the phase forward folding threshold or the phase reverse folding threshold, the initial differential propagation phase shift of the next distance gate is adjusted and the jump amplitude is recalculated, and the abnormal measurement is replaced by the effective measurement of the adjacent distance gate. Numerical range normalization: If the initial differential propagation phase shift is greater than 180°, subtract 360°; if it is less than -180°, add 360° to ensure that the corrected differential propagation phase shift is within ±180°.

7. The method for estimating the detection variables of a time-frequency domain dual-polarization radar with staggered pulse repetition intervals according to claim 1, characterized in that, The sliding average smoothing process uses 10 distance gates for the sliding average smoothing process; the sliding median filter smoothing process uses a processing length corresponding to the number of distance gates in the distance range of 300m to 500m to eliminate small fluctuations generated during the processing and ensure that the estimation results are smooth and undistorted.

8. The method for estimating the detection variables of a time-frequency domain dual-polarization radar with staggered pulse repetition intervals according to claim 1, characterized in that, The Discrete Fourier Transform adapts to two working modes: single-pulse repetition interval and staggered-pulse repetition interval. When the radar operates in single-pulse repetition interval mode, the fast Fourier transform is used to process the horizontally polarized complex baseband signal and the vertically polarized complex baseband signal to directly obtain the horizontally polarized power spectral density and the vertically polarized power spectral density. When the radar operates in staggered pulse repetition interval mode, staggered discrete Fourier transform technology is used to process odd pulse groups and even pulse groups separately to ensure the uniformity of the sampling interval, thereby obtaining the horizontal polarization power spectral density and the vertical polarization power spectral density.

9. The method for estimating the detection variables of a time-frequency domain dual-polarization radar with staggered pulse repetition intervals according to claim 8, characterized in that, When using staggered discrete Fourier transform, the image frequency component removal step is also included: For each distance gate, find the location of the strongest power spectral density among the horizontal polarization power spectral density and the vertical polarization power spectral density; Based on the characteristic that the amplitude of the main frequency region is greater than that of the image frequency component, the effective frequency domain range is determined with the position of the strongest power spectral density as the benchmark. The image frequency component caused by zero-padding interpolation is removed to obtain a pure power spectral density.

10. The method for estimating the detection variables of a time-frequency domain dual-polarization radar with staggered pulse repetition intervals according to claim 1, characterized in that, The calculation of the differential reflectivity also includes attenuation correction, channel error correction, and observation error correction: Attenuation correction is performed based on the attenuation coefficient determined by fitting after actual observation; Channel error correction describes the power measurement difference between horizontal and vertical polarization within the normal linear dynamic range of the radar using a quadratic fitting curve determined during factory calibration. The observation error correction is based on the observation error value determined by the field calibration, in order to improve the accuracy of the differential reflectance estimation.

Images