JP2013195189A - Target angle detection device, target angle detection method and guidance device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain higher angle measuring accuracy.SOLUTION: A target angle detection device includes: a radar receiving section; a target detection section; a first data extraction section; an acceleration correction Fourier transformation section; a second data extraction section; and a monopulse angle estimation section. The radar receiving section demodulates by receiving a radar wave reflected by a target over a plurality of coherent integration periods. The target detection section detects the target from a demodulation result of each coherent integration period, and extracts a speed change for the detected target. The first data extraction section extracts sum signals and difference signals from the demodulation result of the plurality of coherent integration periods for the detected target. The acceleration correction Fourier transformation section performs Fourier transformation processing in order to correct the plurality of sum signals and the difference signals on the basis of the speed change. The second data extraction section extracts a sum signal representative value and a difference signal representative value for one target from a Fourier transformation result. The monopulse angle estimation section performs angle estimation by a monopulse from the sum signal representative value and the difference signal representative value.

Description

  Embodiments described herein relate generally to a target angle detection device, a target angle detection method, and a guidance device that detect a weak target angle.

  There is a track-before-detect (TBD) as a small target detection method in the radar. TBD is a general term for a method of detecting a target having a small SNR (signal to noise ratio) that cannot be detected by 1 CPI by judging a plurality of coherent integration period (CPI) radar received signals together. In normal radar target detection, a point with power equal to or higher than the threshold value is detected for each CPI, but in TBD, the threshold value is not detected for the received signal for each CPI, and the demodulated threshold is not detected. The analog signal is used. In many cases, a target motion model is defined, and it is verified whether or not there is a sequence of points that fit relatively well with the target motion model and have higher power over multiple CPIs. TBD aims to detect targets that have a low SNR that cannot be detected by 1CPI, and is therefore a procedure that scans a large number of candidates to find a higher power point that fits the motion model relatively well. Is generally required, and the amount of calculation is generally large.

  The radar outputs various parameters, but the TBD is configured to scan a large number of candidates as described above. Therefore, when the number of parameters increases, the amount of calculation increases correspondingly. Of the parameters detected by the radar, the angle is a parameter with a relatively large error, so TBD is performed with a small number of parameters excluding the angle, such as range and Doppler frequency, and the angle is re-established from the detected target trajectory. The method of extracting is realistic. Since the angle for each CPI has a large error, it is desired to improve the angle accuracy by detecting the angle from the locus point for each CPI and tracking the angles of a plurality of CPIs used for TBD.

  On the other hand, there is a phase monopulse method as one of the angle measurement methods used in radar. This method generates the sum signal (Σ) and difference signal (Δ) of the outputs of the two antennas, basically the difference signal multiplied by the pure imaginary number is divided by the sum signal, then the position of the two antennas. This is a method of calculating a phase difference and detecting a target angle from the calculated phase difference.

S. J. Davey et al., “A comparison of detection performance for several track-before-detect algorithms”, EURASIP Journal on Advances in Signal Processing, January 2008, Article No. 41 Takashi Yoshida, “Revised Radar Technology”, Chapter 10, IEICE, 1996

  By the way, in the monopulse method, independent thermal noise is added to each of Σ and Δ, and division is performed with the noise added. If there is no signal and only noise, an angle that varies around the front of the beam is output. If the target is not in front of the beam and the SNR of Σ and Δ is small, an error is output between the noise and the target, that is, an angle that varies around the angle between the front of the beam and the target angle. Occur. If there is an error bias in the detection angle of 1 CPI, the error bias cannot be removed even if tracking is performed using a large number of such angles. Accordingly, the error reduction effect due to tracking or the like is small, and it becomes difficult to detect the true target angle.

  On the other hand, it is also considered that the angle is calculated after integrating at each stage of Σ and Δ to obtain a gain to improve the SNR. However, the CPI, which is a single measurement period, is often set to have a maximum length that enables coherent integration when detecting such a weak target. Coherent integration is possible if, for example, a pulse Doppler radar is a period in which the target acceleration can be ignored in the Doppler spectrum and the target can be observed in a nearly line spectrum when Fourier transforming multiple pulses. It is. Even if the period is longer than that, the spectrum is blurred due to the target acceleration, and the gain does not increase. Therefore, it is difficult to simply integrate Σ and Δ over a period longer than 1 CPI period.

  An object of the present invention is to alleviate the problem that the error in the angle measurement result due to monopulse has a bias that cannot be eliminated by tracking or the like at low SNR, and to achieve higher angle measurement accuracy, target angle detection method, target angle detection method, and guidance device Is to provide.

  According to the embodiment, the target angle detection apparatus includes a radar receiver, a target detector, a first data extraction unit, an acceleration correction Fourier transform unit, a second data extraction unit, and a monopulse angle estimation unit. Is provided. The radar receiver has two antennas for use in monopulse angle measurement, receives and demodulates the radar wave reflected by the target over a plurality of coherent integration periods, and outputs each of the coherent integration periods. The demodulation result is output. The target detection unit detects a target from the demodulation result of each coherent integration period output from the radar receiver, and extracts a speed change in the plurality of coherent integration periods for the detected target. The first data extraction unit extracts a sum signal and a difference signal of the two antennas for the detected target from the demodulation results in a plurality of coherent integration periods, respectively. The acceleration correction Fourier transform unit performs a Fourier transform process to correct the plurality of sum signals and difference signals extracted by the first data extraction unit based on the speed change extracted by the target detection unit. The second data extraction unit extracts one sum signal representative value and one difference signal representative value per target from the Fourier transform result. The monopulse angle estimation unit performs monopulse angle estimation from the sum signal representative value and the difference signal representative value.

The block diagram which shows the structure of the target angle detection apparatus of one Embodiment. The figure which shows the example which simulated the error bias absolute value in one Embodiment. The figure which shows the spectrum of the result of having carried out the fast Fourier transformation of (SIGMA) and (DELTA) for 50 CPI as a comparative example of one Embodiment. The figure for demonstrating the phase which the head pulse of each CPI has. The block diagram which shows the example of the acceleration correction | amendment Fourier-transform part which implements the 1st method in one Embodiment. The figure which shows the spectrum of the result of carrying out the fast Fourier transform after correcting a phase in one Embodiment. The block diagram which shows the example of the acceleration correction | amendment Fourier-transform part which implements the 2nd method in one Embodiment. The block diagram which shows the form of (Sigma) * delta representative value extraction part in one Embodiment. The block diagram which shows the other example of an acceleration correction | amendment Fourier-transform part in one Embodiment. The block diagram which shows the other example of (SIGMA. * (DELTA) representative value extraction part) in one Embodiment. FIG. 4 is a diagram illustrating error RMSE and error bias performance in one embodiment. The block diagram which shows the example of a target detection part in one Embodiment. The figure for demonstrating TBD. The figure for demonstrating TBD similarly. The figure which shows one Embodiment of the guidance apparatus of a flying body. The flowchart which shows the control processing procedure of the target angle detection apparatus in other embodiment.

  Hereinafter, embodiments will be described with reference to the drawings.

  FIG. 1 is a block diagram illustrating a configuration of a target angle detection device according to an embodiment. In the target angle detection device 1, a radar wave reflected by a target, not shown, or a radio wave output by the target is captured via the antenna 2 and received by the radar receiver 3. The antenna 2 effectively has two antennas in each direction so that monopulse angle measurement is possible. In FIG. 1, one antenna is shown for the sake of simplicity. Having two antennas effectively is not only when there are really only two antenna elements, but also when one of the monopulse antennas forms a phased array antenna with multiple antenna elements. Including. Moreover, you may have another antenna which is not utilized for monopulse angle measurement. In addition, the fact that there are effectively two antennas per direction means that if the monopulse angle measurement direction is the elevation angle and the azimuth, two antennas are effective in the horizontal direction and two antennas are effective in the vertical direction. For example, in the form of a phased array antenna.

  The radar receiver 3 converts radio frequency (RF) radio waves captured by the antenna 2 into baseband, performs appropriate demodulation according to the radio wave format, and outputs a demodulation result. For example, in the case of a chirped pulse type pulse Doppler radar, first, individual pulses are subjected to pulse compression, the pulses in the CPI are Fourier transformed for each range bin, and the result is output. The radar receiver 3 outputs a demodulation result over a plurality of CPIs. The radar receiver 3 outputs both the Σ demodulation result and the Δ demodulation result used in the monopulse, or outputs at least a demodulation result sufficient to extract Σ and Δ at the subsequent stage. For example, the demodulation result of each antenna is output.

  The demodulation result output from the radar receiver 3 is input to the target detector 4. The target detector 4 detects a desired target from the demodulation result of the radar receiver 3. Furthermore, the target detection unit 4 detects the speed of the detected target at each CPI, that is, how the target speed changes over a plurality of CPIs.

  Next, the CPI unit Σ · Δ extraction unit 6 inputs the target detection result of the target detection unit 4 together with the demodulation result from the radar reception unit 3. The CPI unit Σ · Δ extraction unit 6 extracts Σ and Δ of two antennas used for monopulses from the corresponding CPI demodulation result at the point where the target is estimated by the target detection unit 4 for each CPI. For one target, Σ and Δ are extracted one by one for 1 CPI. For example, in the case of a chirp pulse type pulse Doppler radar, one range Doppler map for each Σ and Δ is obtained for one CPI, and one Σ and Δ for one target is extracted therefrom. That is, Σ and Δ in a state where the integral gain by demodulation is taken are extracted.

  Σ and Δ of each extracted CPI are supplied to the acceleration correction Fourier transform unit 7. Further, the acceleration correction Fourier transform unit 7 is inputted with the speed at each CPI of the target detected from the target detection unit 4, that is, the speed change. The acceleration correction Fourier transform unit 7 performs Fourier transform so as to correct the acceleration of Σ and Δ of each CPI using the input speed change. Details will be described later.

  The result of Fourier transform in this way is supplied to the Σ · Δ representative value extraction unit 9. The Σ · Δ representative value extraction unit 9 extracts the values of Σ and Δ as one representative value per target from the input Fourier transform result. The extraction method will be described later.

  The extracted representative values of Σ and Δ are supplied to the monopulse angle estimation unit 10. The monopulse angle estimation unit 10 performs monopulse angle measurement processing using the representative values of Σ and Δ, and outputs target angle information viewed from the antenna 2 obtained thereby.

  In this way, with respect to the low SNR target, the influence of the target acceleration is suppressed, and the integral gain is secured over a plurality of CPIs at the stages of Σ and Δ to improve the SNR, and then monopulse angle measurement is performed. It becomes possible. Since the SNR of Σ and Δ is improved, the error bias is suppressed and the angle measurement accuracy is improved.

The principle of this embodiment will be described below. First, the monopulse angle measurement formula will be described below. The tangent Φ corresponding to half the phase difference of the antenna 2 is very simply calculated by the following equation.

The equation is based on the assumption that Δ and Σ are orthogonal on the complex plane, and it is usually difficult to guarantee the orthogonality between Δ and Σ without noise. In particular, in a state where noise is added at a low SNR, the orthogonality between Δ and Σ is low. Non-orthogonal components may be simply excluded, but a frequently used equation is as follows.

This is a method of projecting Δ on an axis orthogonal to Σ. Since the purpose is to obtain the phase difference between the two antennas, other formulas exist but are omitted. Hereinafter, in order to simplify the description, the description is based on the formula (1), but the simulation result and the like are based on the formula (2). Φ obtained in this way is input to the following equation to obtain an angle θ [rad].

Where φ _B is the beam angle, λ is the carrier frequency wavelength, and d is the antenna spacing. The beam angle is generated at an angle other than 0 degrees when the beam direction can be controlled, for example, each monopulse antenna is a phased array antenna.

Next, consider a case where the SNR of the target reflected wave is low. In monopulse angle measurement, in many cases, Δ and Σ are combined at the antenna output end and then amplified by an amplifier. Therefore, thermal noise is added in the state of Δ or Σ. When noise is included, equation (1) is as follows.

n _d is the thermal noise added to Δ, and n _s is the thermal noise added to Σ. Originally, jΔ / Σ corresponds to the correct target angle, and we want to find this value. However, the effect of noise cannot be ignored at low SNR where the noise power is comparable to the signal power. If only noise is present, jn _d / n _s is distributed around the front of the beam. At a low SNR, each of these noises has a shape such that jΔ and Σ are added to each other, and the distribution is such that the front of the beam and an intermediate value of the target angle are varied. As a result, a bias occurs in the error, and even if integration processing such as averaging is performed after the angle is obtained, the measurement angle does not approach the target true angle.

  In FIG. 2A, when the antenna interval is 3.5λ, the target is 5 degrees, and the front of the beam is 3 degrees, the SNR per antenna is variously changed, and the target angle error RMSE is measured. This is an example in which an error bias absolute value, which is a deviation of the average value of the measured angles from the correct target angle, is simulated. It can be seen that the error bias starts to exceed 0.1 degree when the SNR per antenna falls below 5 dB, and the error bias of about 1 degree occurs at -5 dB.

  Similarly, FIG. 2B shows the RMSE and error bias when the SNR per antenna is fixed at -1 dB, the target angle is fixed at 5 degrees, and the beam angle is changed on the horizontal axis. It can be seen that the error bias decreases near the beam angle of 5 degrees, and the error bias increases and the RMSE increases with increasing distance from the beam angle.

  FIG. 2A shows that the error bias can be suppressed if the SNR increases. That is, for example, even if the SNR per antenna is -5 dB, the error bias can be reduced to 0.1 degrees if the SNR can be increased to 5 dB by giving a gain of 10 dB by some method. For example, when there is a sample of -5 dB, a gain of 10 dB can be obtained if 10 samples of the same condition can be integrated coherently. However, as described above, this gain needs to be taken before Δ is divided by Σ.

  In the present embodiment, one Δ and Σ are extracted per target for each CPI, and an integral gain is obtained by subjecting Δ and Σ of a plurality of CPIs to Fourier transform. By using ΔC and Σ of 1 CPI as one sample of Fourier transform, and at the time of Fourier transform, the acceleration is corrected by using the speed change of the target detection result so that an integral gain can be obtained even during a period longer than 1 CPI. To.

  These will be described below. FIG. 3A shows that when the target acceleration is completely fixed at 0, pulse Doppler radar detects the target Δ and Σ over 50 CPI, and Σ and Δ for 50 CPI are fast Fourier transform respectively. It is the spectrum of the result of conversion. The SNR per 1 CPI / antenna is 1 dB, 1 CPI is 10 ms, and the carrier frequency is the Ku band. If the acceleration is completely zero, the target appears as a line spectrum for both Σ and Δ, and it can be seen that an integral gain can be obtained by Fourier transform.

On the other hand, FIG. 3B shows a spectrum of a Fourier transform result for 50 CPI when the target has an acceleration of about 3 m / s ^{2 under} substantially the same conditions as in FIG. It can be seen that even if a small acceleration is applied, no integral gain can be obtained by Fourier transform, and there is no line spectrum. In reality, it is difficult to assume that the target acceleration is completely zero over a plurality of CPIs, and it can be seen that an integral gain cannot be obtained even if a Fourier transform is normally performed over a plurality of CPIs.

  In the pulse Doppler radar, a plurality of pulses in one CPI are Fourier transformed. The resulting frequency bin spacing of the spectrum is 1 / CPI. When a sample for a plurality of CPI is Fourier transformed with one sample per CPI, the range from end to end of the spectrum is 1 when there is no extra elapsed time between CPIs and adjacent CPIs are continuous. / CPI. That is, the entire range of the spectrum of FIG. 3 is equal to the range of 1 bin of the spectrum of 1 CPI pulse Doppler radar, and is very narrow. As described above, assuming that the period of 1 CPI is the longest period for which the target can barely maintain the line spectrum, in practice, the spectrum is often spread over one bin of the spectrum for 1 CPI, As shown in FIG. 3B, it is natural that a line spectrum cannot be observed with a spectrum that observes the contents of one bin.

  In the present embodiment, the target speed detected by the target detection unit 4 and the change in speed are corrected to obtain a spectrum close to a line spectrum.

  In normal pulse Doppler radar, the purpose is to detect the target from the range Doppler map, so it is unclear what speed the target will have at the time of the Fourier transform. Although the change cannot be corrected, in this embodiment, the target detection is performed in advance by the target detection unit 4, and the purpose of the Fourier transform in the acceleration correction Fourier transform unit 7 is not target detection but angle detection. Therefore, such a procedure is possible.

  First, what kind of correction is necessary will be described, and then the correction method will be described.

FIG. 4 is a diagram for explaining the phase of the leading pulse of each CPI. The horizontal axis represents time, and the thick arrow indicates the time taken when the radar wave of each CPI is reflected back and forth by the target. Since the target gradually approaches the radar, the round trip time becomes shorter as the CPI number advances. The formula indicated by the thin arrows toward each thick arrow is the phase of the radar wave of each CPI. The phase changes from moment to moment depending on the pulse number in each CPI, the time in the pulse, and the like, but by performing appropriate demodulation, the phase represented by the value shown in the figure can be obtained. ω ₀ is the carrier frequency (ignoring Doppler frequency), R ₀ is the target range at CPI number 0, c is the speed of light, v ₀ is the target speed at CPI number 0, and Δv _m is the speed at CPI number m the difference between the v _0, T ₂ is CPI length, k is the CPI number. As the CPI progresses, the target moves and the round trip time becomes shorter. If the target speed is constant, there is a 2v ₀ kT ₂ / c round trip time change from 0 to k, but there is a change in the target speed,

A term corresponding to the change in distance is added. Because of this term, the spectrum spreads even when the usual Fourier transform is applied.

  In the present embodiment, two methods are used to correct the term of equation (5). One is a method of removing this term before the Fourier transform, and the other is a method of correcting the Fourier transform kernel in accordance with the speed change at the time of the Fourier transform.

In the first method, when there is no term in equation (5), the phase at CPI number k is as follows.

If equidistant phases are obtained with respect to k and the amplitude is constant, a line spectrum is obtained by ordinary Fourier transform (Fourier series transform). Therefore, Fourier transform is performed after removing the corresponding term of equation (5) from each of Σ and Δ of each CPI. Delta] v _m has been detected by the target detection unit 4, at the time of Fourier transform to remove the phase term corresponding to utilize the detection result (5). Details of the target detection unit 4 will be described later.

FIG. 5 shows a form of the acceleration correction Fourier transform unit 7 according to the first method.
The acceleration correction Fourier transform unit 7 includes a speed change correction unit 5 and a fast Fourier transform unit 8. The acceleration correction Fourier transform unit 7 receives CPI Σ and Δ used for Fourier transform from the CPI unit Σ · Δ extraction unit 6. At the same time, the target speed detected by each target CPI is notified from the target detection unit 4.

  First, the speed change correction unit 5 removes the phase related to the equation (5) for each Σ and Δ. Subsequently, in the fast Fourier transform unit 8, the result is Fourier transformed for each of Σ and Δ. In the fast Fourier transform, a spectrum can be obtained by a single transformation, so that a spectrum for each of Σ and Δ can be obtained.

  FIG. 6 is a spectrum resulting from fast Fourier transform after correcting the phase corresponding to equation (5) for the data in FIG. Here, it is assumed that the correct speed can be detected by the target detection unit 4. A sharp line spectrum similar to that shown in FIG. 3A is obtained, and it can be seen that an integral gain can be obtained by Fourier transform if the speed change is corrected.

  In the first method, the fast Fourier transform is used, so that a spectrum is obtained at one time. However, if the accuracy of the speed of each CPI obtained by the target detection unit 4 is high, the Fourier transform is performed using the speed transform. It may be a discrete Fourier transform at one frequency specifying (Doppler frequency). In this case, a spectrum is not obtained, and only a complex value of any frequency is obtained, and this value is used as it is as a Σ representative value and a Δ representative value.

  Another phase correction method including the equation (5) is to change the kernel of the Fourier transform in accordance with the phase of the radar wave of each CPI shown in FIG. Is the method.

The Fourier transform is generally expressed by the following equation.

Where x (t) is a function that undergoes Fourier transform. In signal processing such as pulse Doppler radar, the time t in equation (7) is discretized and processed. In most cases, the time t is discretized at regular intervals. However, from the viewpoint of discretizing the equation (7), the discretization does not necessarily have to be at regular intervals. On the other hand, the reason why the line spectrum cannot be obtained well even if the Fourier transform is performed with the phase of the CPI sequence in FIG. 4 as it is explained in the equation (5), but the reception timing of the radar wave of each CPI by the equation (5). The reason for the unequal spacing is that the change in the range due to acceleration is reflected. Therefore, it can be expressed not in the form shown in FIG. 4 but in the form exp (−jω ₀ 2R (t) / c), for example. The expression (7) can be further regarded as an expression that correlates the integrand x (t) with a sine wave having an angular frequency of ω, so it matches the acquisition timing of R (t). If the timing for discretizing t is changed, the data sequence having the phase shown in FIG. 4 can be directly Fourier transformed. The part corresponding to exp (-jωt) in equation (7) is called the conversion kernel. In the second method, the Fourier transform kernel is set as follows.

As can be seen from the equation, the target detection unit 4, information on Delta] v _m and v _0, that is, it is necessary to obtain the speed of each CPI used in the Fourier transform. Also, since fast Fourier transform cannot be applied to this Fourier transform, it is necessary to apply discrete Fourier transform (DFT). Similarly to the first method, if the accuracy of the velocity information is high, a single DFT for one v ₀ is applied to Σ and Δ, respectively, and the resulting complex values are represented as Σ representative values, Δ Can be an extractor.

However, it is unlikely that the speed information obtained by the target detection unit 4 is accurate to an accuracy that is a fraction of one bin of a pulse Doppler spectrum of 1 CPI. Therefore, it is preferable to obtain a spectrum similar to that obtained by fast Fourier transform and detect the peak. Since it is a DFT, v ₀ is variously changed and Fourier transformed to obtain a spectrum.

As described above, since the sample interval is the CPI interval, the frequency range of the obtained spectrum is 1 / T ₂ , and frequencies higher than that appear in the remainder frequency divided by 1 / T ₂ . Therefore, the value of v ₀ is not necessary at all, and 1 / T ₂ is divided by the number N of samples to be Fourier-transformed or larger, and ω ₀ = 2πf ₀ is set, and f ₀ 2vT ₂ / c is changed from 0 to 1 / v may be set so that 1 / T ₂ / N intervals or less until T _2. By doing so, a spectrum equivalent to the first method can be obtained by the second method. The reason why the interval is 1 / T ₂ / N or less is that 1 / T ₂ / N is the maximum interval at which information is not lost. This is because even if it is a spectrum, it may not appear in the spectrum.

FIG. 7 shows a form of the acceleration correction Fourier transform unit 7 that performs the second method.
The acceleration correction Fourier transform unit 7 includes a discrete Fourier transform unit 11, a kernel setting unit 12, and a frequency designation unit 13.

Information regarding the speed input from the target detection unit 4 is input to the kernel setting unit 12. Here, a kernel like equation (8) synchronized with the target speed change is set and input to the discrete Fourier transform unit 11. The discrete Fourier transform unit 11 receives Σ and Δ of CPI used for Fourier transform, which are input from the CPI unit Σ · Δ extraction unit 6. The frequency specifying section 13 specifies the frequency to be used in the Fourier transform as bottles obtained spectrum becomes 1 / T ₂ / N or less of the interval from 0 to 1 / T _2. The spectrum of Σ and Δ obtained in this way is output from the acceleration correction Fourier transform unit 7.

  One of the features of this embodiment is that processing similar to that performed by pulse Doppler radar is performed in units of CPI, not in units of pulses. By performing in CPI units, the resulting spectrum is folded back in very fine frequency units. If this is a pulse Doppler radar, the Doppler frequency, that is, the speed ambiguity is too large, and the speed detection performance is low. Damaged. However, in this embodiment, the speed and the like are detected in advance by the target detection unit 4, and Fourier transform is performed in order to obtain an integral gain for improving the angle detection accuracy. By performing Fourier transform in units of CPI so that the spectrum is folded back in fine units, there is an advantage that the spectrum range to be investigated can be narrowed. For example, the correction of acceleration is performed by returning to the pulse unit instead of the CPI unit as in the present embodiment, and the Σ representative value and Δ representative value can be detected from the peak of the obtained spectrum. Such a method increases the computational complexity by expanding the spectrum. In addition, since the number of bins in the spectrum increases, with low SNR targets, the probability of mistaking a noise peak as the target increases and performance degrades. Therefore, by performing the Fourier transform on the CPI unit sample, not only the calculation amount but also the performance can be improved.

  In the embodiment of the acceleration correction Fourier transform unit 7 according to the first and second methods, a spectrum is obtained instead of a Fourier transform result for a single velocity, and a Σ representative value and a Δ representative value are detected from the peak. An example of an embodiment of the Σ · Δ representative value extraction unit 9 corresponding to this form is shown in FIG.

  In both forms of the first and second methods, two spectra are obtained: a Σ spectrum and a Δ spectrum. In FIGS. 3 and 6, since the target is closer to the front side than the 3 dB width of the beam, the value of Σ is larger. However, when the target is at the end of the beam, Δ is larger. . In monopulse angle measurement, Σ and Δ must always use the same bin value. If the bin number is determined from the peak of the spectrum of Σ and the representative value of Δ is taken from the spectrum of Δ of the same bin number, the bin where the target exists is found when the power of Σ is small. There is a possibility of mistakes.

  Therefore, first, the spectrum maximum value detection unit 14 obtains a bin number and a peak value that give a power peak of each of the Σ spectrum and Δ spectrum. Next, the bin number determination unit 15 compares these peak powers and selects the bin number of the peak of the larger spectrum. The selected bin number is notified to the complex value extraction unit 16. The complex value extraction unit 16 extracts the notified complex value of the bin number from the Σ and Δ spectra, and outputs these as the Σ representative value and the Δ representative value.

  By doing so, it is possible to reduce the probability of mistaken bins for extracting Σ · Δ representative values, regardless of the position of the target in the beam.

  In the case where the acceleration correction Fourier transform unit 7 performs only a Fourier transform for a single speed and a single speed change, the Σ / Δ representative value extraction unit 9 outputs the output of the acceleration correction Fourier transform unit 7. Is output as is.

In the first and second methods described above, it is assumed that Δv _m is relatively accurate. However, in practice, accurate detection cannot be expected when Δv _m is also low in SNR. Therefore, as a modification of the present embodiment, the acceleration is swung in a certain range, and the acceleration that maximizes the maximum power of the Fourier transform result is selected.

FIG. 9 shows an example of the acceleration correction Fourier transform unit 7 that implements the above modification. In FIG. 9, the result of the target detection unit 4 is once input to the speed change range designation unit 17. In the speed change range designating unit 17, the speed change, i.e. with respect to Delta] v _m, to some set values around the value target detection unit 4 is output. For example, the Delta] v _m of the target detecting section 4 is output, a sequence obtained by multiplying a constant coefficient, set by changing several sizes of coefficients, and the like. The value of Delta] v _m of the set plurality of streams is notified to the acceleration correction Fourier transform unit 7. The acceleration correction Fourier transform unit 7 may employ either the first method or the second method. Then, the acceleration correction Fourier transform unit 7 outputs a spectrum of Σ and Δ that has been subjected to correction in each series and subjected to Fourier transform. Therefore, the acceleration correction Fourier transform unit 7 outputs a plurality of spectra for each of Σ and Δ. This result is input to the Σ · Δ representative value extraction unit 9.

  FIG. 10 shows an example of the Σ · Δ representative value extraction unit 9 that implements the above modification. As shown in FIG. 10, in the Σ / Δ representative value extracting unit 9, the spectrum maximum value detecting unit 14 first extracts the maximum power value and bin number of each of the Σ and Δ spectra of the plurality of speed changes. , Output to the speed change / bin number determination unit 18. Next, the speed change / bin number determination unit 18 selects a spectrum of Σ or Δ of the speed change having the largest power maximum value, extracts the speed change and the bin number of the power maximum value, and outputs a complex value extraction unit. 16 is notified. The complex value extraction unit 16 extracts the complex value of the notified bin number from the notified spectrum of Σ and Δ of the speed change, and outputs these as a Σ representative value and a Δ representative value.

Incidentally, depending on the configuration of the target detection unit 4, changes in Delta] v _m varies relatively greatly CPI basis rather than gradual. CPI is as small as about 10 ms, if the target and the radar has a certain mass, possibly rattle Delta] v _m within a short time is small, large fluctuations in CPI unit backlash due performance of the detection method employed by the target detecting section 4 There is a high possibility that it is dead. In particular, in a method that can detect a target locus only in bin units as will be described later, when the bin interval is rough, Δv _m rattles.

As can be seen from equation 4, the dimension of the correction of the phase range, i.e., to be done by the dimension of the integration of Delta] v _m, looseness temporary in Delta] v _m, but it is later than the phase correction of the CPI The influence spreads in the form of saying that it affects. Therefore, when the credibility of the rattling of the speed output from the target detection unit 4 is low, such as when the target detection method is in bin units, smoothing may be performed in advance so that the change in Δv _m becomes gentle. . Simply, fitting the speed to a polynomial with a least square error criterion, etc. At that time, the degree of the polynomial should be a gentle curve of 2nd order or 3rd order at most. A linear approximation with degree 1 is sufficient.

  Of course, if the target period of Fourier transform by a plurality of CPIs becomes long, the target may actually take a complicated movement that cannot be approximated by a low-order polynomial. However, it is very difficult to perform correction so that an integral gain can be obtained following such a complicated change in acceleration, as described above, since a small error is integrated later and echoes.

  Therefore, in the present embodiment, the CPI range to be Fourier-transformed by the acceleration correction Fourier transform unit 7 is not set so wide. That is, the target detection unit 4 uses a plurality of CPI demodulation results for target detection, and performs Fourier transform using Σ and Δ of CPI in a range equal to or less than the range of CPI used at that time.

  FIGS. 11A and 11C show simulation results for verifying the performance of the present embodiment under the same conditions as FIG. This is the RMSE and error bias of 100 trials as a result of obtaining a spectrum for a plurality of speed changes with slightly different acceleration around the correct speed change, detecting the maximum value, and performing monopulse angle measurement. The horizontal axis is the number of CPI used for Fourier transform. In addition, the range in which the speed is changed is a range in which the speed is shifted by ± 1 bin by changing the Doppler bin of the range-Doppler map in 1 CPI in the period of 50 CPI, and the range is divided into 20 to correspond to each speed change spectrum. Ask for.

  The solid lines in FIGS. 11A and 11C indicate the speed change before and after the speed change giving the maximum power when the spectrum of the speed change giving the maximum power is selected and the Σ · Δ representative value is extracted. This is a case where it is intentionally selected to be a Σ · Δ representative value. As long as the number of CPI used for Fourier transform is small, there is no problem in the performance of the correction even if there is an error in the speed change used for correction. Both error bias and performance deteriorate. When the number of CPI is more than a certain level, the performance is deteriorated as compared with the case where the number of CPI is small.

  Actually, it is desirable that the number of CPIs used for Fourier transform is not so large considering that the true value of the speed change does not completely coincide with the detection result or the gentle polynomial.

  According to FIG. 2, if the SNR per antenna is about 5 dB, the error bias can be made sufficiently small. Even if TBD is performed by the target detection unit 4, the SNR that can be detected by the target has a lower limit in most cases, and it is only necessary to obtain a gain sufficient to increase the SNR from the lower limit to about 5 dB at most. For example, if the lower limit of the SNR at the time of target detection is -5 dB, a gain of 10 dB, that is, 10 CPI, or a CPI number about twice that is sufficient in consideration of the case where correction is not possible is sufficient, and a very large number Is not necessary.

  In the condition of FIG. 11, the average value of the angle measurement error bias in CPI units is about 0.3 degrees, but according to FIG. 11C, the error bias absolute value is less than 0.1 degrees at about 10 CPI. It can be seen that the error bias is sufficiently suppressed.

  FIGS. 11B and 11D show the error RMSE and error bias when the monopulse angle measurement is performed from the peak of the maximum power of the spectrum by simply performing Fourier transform without correcting the acceleration. Since the error RMSE in CPI units is about 2 degrees, the RMSE is slightly improved as compared with nothing, but the performance is clearly degraded as compared with the case of acceleration correction. In addition, the error bias is often degraded from the error bias in CPI units. This is because when the acceleration is not corrected, the spectrum is as shown in FIG. 3B, and even if the power peak is selected, it is not always the SNR maximum bin.

  Next, the purpose of the target detection unit 4 is to determine a point where the target is estimated to exist based on the demodulation result of each CPI, and extract Σ and Δ for calculating the angle to the CPI unit Σ / Δ extraction unit 6. And instructing the acceleration correction Fourier transform unit 7 of the target detection speed in each CPI used for Fourier transform.

  The target detection unit 4 detects the target from the demodulation result output from the radar reception unit 3 as described above. There are various detection methods. The target may be detected from only a part of a plurality of CPI demodulation results, and the target state of other CPI may be estimated from the result, or the target is detected using all of them. May be. In any case, for the CPI used in the acceleration correction Fourier transform unit 7, the point and speed where the target is estimated to exist are estimated, and Δ and Σ can be extracted therefrom.

  FIG. 12 is an example of the target detection unit 4, in which the target detection unit 4 performs TBD to detect the target. First, the target detection unit 4 stores the demodulation results of a plurality of CPIs in the demodulation result storage unit 19, and uses the plurality of demodulation results stored here to execute TBD in the TBD execution unit 20. The point and the speed at which the target is estimated to exist in the demodulation result are output.

  13 and 14 are diagrams for explaining the TBD. In each figure, the horizontal axis represents the range, the vertical axis represents the Doppler frequency, and the Σ demodulation result of the pulse Doppler radar is represented by a color map. The darker the color, the stronger the power at that point. The target is at 5 degrees and the beam is at 3 degrees. FIG. 13 shows an example in which the SNR per target antenna is 1 dB, and the target gradually moves from (a) to (j). The target position is indicated by an arrow in the figure. Although it is difficult to distinguish, the point where the target exists is slightly darker than the surroundings. FIG. 14 is a map corresponding to FIG. 13F when the target SNR is 14 dB. If there is a difference between the target and the noise power of about 14 in FIG. 14, it is possible to detect the target from one map. However, in the state as shown in FIG. 13, it is difficult to correctly detect the target from one map. There is a need for TBD processing that defines a target motion model, considers multiple maps, and searches for a trajectory that matches the motion model. By applying TBD, it is possible to detect the target trajectory even if the target SNR is low, and as a result, it is possible to extract Σ and Δ for calculating the angle along the trajectory with a high probability that the target exists As a result, the angle accuracy is improved.

  Although there are various TBD methods, the operation of the present embodiment is not dependent on the specific TBD method itself, and is therefore omitted.

  In many cases, the calculation amount of TBD is very large. Also, the amount of calculation increases with the number of map dimensions for performing TBD. In this embodiment, a pulse Doppler radar is assumed as a radar. The demodulation result output by the pulse Doppler radar is basically information on the range and Doppler (speed), but in the case of a configuration having an angle measuring function, an angle can also be output. In many documents, when TBD is performed, a map in which a target position is converted into a Cartesian coordinate system is created from two directions or one direction of an angle and a range, and TBD is performed. However, the angle measurement accuracy is low compared to the range detection accuracy, and is particularly low due to the above-mentioned error bias problem and the problem of glint noise that occurs when the target has multiple reflection points. In SNR, the reliability of the target angle is low. Therefore, in the case of the pulse Doppler radar, the TBD is performed with only the dimensions of the range and the Doppler, so that the calculation amount is small and the detection accuracy is high. In the present embodiment, since speed information for correcting the speed change of the CPI used in the acceleration correction Fourier transform unit 7 is necessary, speed detection can be easily performed by performing TBD with parameters including the Doppler frequency. Become.

  Of course, in the present embodiment, a radar other than the pulse Doppler radar, for example, a synthetic band radar or FM-CW radar may be used. However, the TBD step does not include an angle, and includes a Doppler frequency. It is desirable to use parameters.

  The CPI unit Σ / Δ extraction unit 6 receives information on the point where the target is estimated to exist from the target detection unit 4, and extracts Σ and Δ at that point from the demodulation result of each CPI. As for the granularity of the trajectory output by the TBD, there are roughly one that outputs in units of bins of the map of the demodulation results (for example, Viterbi) and one that can output to units smaller than the bins (for example, Bayes).

  When outputting in bin units, the CPI unit Σ / Δ extracting unit 6 extracts the Σ and Δ values of the bins indicated by the demodulation results of each CPI. If the demodulation result output from the radar receiver 3 takes the form of a Σ map or a Δ map, the instructed bin value may be taken as it is. In the case of outputting the demodulation results of two antennas, it is only necessary to generate Σ from the sum of the bin values and Δ from the difference.

  When the TBD outputs to a unit finer than the bin, the points estimated to exist in each CPI may be rounded to the bin unit, and the bin to be extracted may be determined and processed in the same manner. However, if the data before demodulation of each CPI is held, it may be demodulated only at that point so that Σ and Δ can be extracted at the point instructed by the target detection unit 4. For example, if it is a chirped pulse type pulse Doppler radar, if it holds the spectrum of each pulse, the pulse representative value of the specified range is extracted from the pulse spectrum by DFT, and from the pulse representative value sequence The value of the designated Doppler frequency (speed) is also extracted by DFT. If these are performed for Σ and Δ, respectively, even if the locus is finer than the bin unit, Σ and Δ of the fine points can be extracted. On average, the SNRs of Σ and Δ become smaller as they deviate from the point of the correct locus. By rounding to the bin unit, SNR degradation of several dB may occur. Therefore, by obtaining Σ and Δ according to the instruction of the locus finer than the bin unit, it is expected that the angle measurement accuracy is improved to some extent.

  In the methods shown in FIGS. 9 and 10, the target speed and acceleration can be detected with higher accuracy than TBD by using the spectrum obtained as a result of the acceleration correction Fourier transform. Accordingly, the detected speed / acceleration results may be reflected in the TBD, and these may be set to known values to execute the TBD again to improve the target detection performance. Alternatively, the accuracy of the target detection parameter may be increased by replacing the speed and acceleration of the TBD result with the values detected in the present embodiment.

  FIG. 15 is a block diagram showing an embodiment of the flying object guiding apparatus 21. The guidance device 21 includes a target angle detection device 1 and a guidance signal generation unit 22. The guidance signal generation unit 22 further adds a target range, speed, and the like to the target angle information output from the target angle detection device 1 of the present embodiment, if necessary, to the guidance signal generation unit 22. Output. The guidance signal generation unit 22 generates a guidance signal for guiding the flying object from the input angle or the like, and controls a flying body driving unit (not shown).

  By doing this, even if the target SNR is weak and an error bias occurs in the angle with monopulse angle measurement, the influence of the error bias when the accuracy of the angle is increased by integrating the results of multiple CPIs. This makes it possible to control flying objects with reduced target tracking performance.

  As described above, according to the above-described embodiment, before the target angle is detected in the target angle detection device 1, the target detection unit 4 performs target detection with other parameters, and the target speed change is detected from the detection result. The sum signal and the difference signal are detected even in a period exceeding 1 CPI by performing Fourier transform so that the acceleration correction Fourier transform unit 7 detects and corrects the sum signal and difference signal of each CPI based on the detected speed change. Is obtained. Further, instead of performing Fourier transform from the pulse stage, Fourier transform is performed on samples in CPI units, thereby reducing the amount of calculation and improving performance. As a result, the effect of error bias generated when monopulse angle measurement is performed at a low SNR is reduced, and higher angle measurement accuracy can be obtained.

  In the target angle detector 1 of the above embodiment, the radar receiver 3, the target detector 4, the CPI unit Σ / Δ extraction unit 6, the acceleration correction Fourier transform unit 7, and the Σ / Δ representative value extraction unit. 9 and the monopulse angle estimator 10 have been described as hardware examples. The radar receiver 3, the target detector 4, the CPI unit Σ / Δ extractor 6, the acceleration correction Fourier transform unit 7, The processing of the Δ representative value extraction unit 9 and the monopulse angle estimation unit 10 can be processed by software using a computer program.

  FIG. 16 is a flowchart showing a control processing procedure of the target angle detection apparatus 1.

  First, the target angle detection device 1 converts a radio frequency (RF) radio wave captured by the antenna 2 into a baseband, performs an appropriate demodulation corresponding to the radio wave format, and performs demodulation results over a plurality of CPIs. Is obtained (step ST16a). Then, the target angle detection device 1 detects a target locus from the demodulation results of a plurality of CPIs by track-before-detect (step ST16b), and extracts Σ and Δ of each point of the target locus (step ST16c). Then, a correction value corresponding to the target speed change in each CPI is designated (step ST16d), and the accelerations of Σ and Δ of the plurality of CPIs are corrected based on the designated correction value (step ST16e), and the corrected Σ And Δ are subjected to fast Fourier transform (step ST16f).

  Subsequently, the target angle detection device 1 determines whether or not the processing has been completed for all designated speed changes (step ST16g). Here, if the process is not completed (no), the target angle detection device 1 repeatedly executes the processes of the above-described steps ST16d to ST16f.

  On the other hand, if the processing is completed (yes), the target angle detection device 1 detects the power maximum value of the fast Fourier transform result spectrum of Σ and Δ of each CPI (step ST16h), and most of these power maximum values. A bin number corresponding to a large power maximum value is selected (step ST16i). Then, the target angle detection apparatus 1 extracts the complex value of the fast Fourier transform result of the selected correction value and Σ of the bin number as the Σ representative value, and the complex of the fast Fourier transform result of Δ of the same correction value and the same bin number. The value is extracted as a Δ representative value (step ST16j). Thereafter, the target angle detection device 1 performs monopulse angle measurement processing using the representative values of Σ and Δ, and outputs target angle information viewed from the antenna 2 (step ST16k).

  Furthermore, the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

  DESCRIPTION OF SYMBOLS 1 ... Target angle detection apparatus, 2 ... Antenna, 3 ... Radar receiving part, 4 ... Target detection part, 6 ... CPI unit (SIGMA) * delta extraction part, 7 ... Acceleration correction Fourier-transform part, 8 ... Fast Fourier-transform part, 9 ... Σ · Δ representative value extraction unit, 10 ... monopulse angle estimation unit, 11 ... discrete Fourier transform unit, 12 ... kernel setting unit, 13 ... frequency designation unit, 14 ... spectrum maximum value detection unit, 15 ... bin number determination unit, 16 DESCRIPTION OF SYMBOLS ... Complex value extraction part, 17 ... Speed change range designation | designated part, 18 ... Speed change and bin number determination part, 19 ... Demodulation result storage part, 20 ... TBD execution part, 21 ... Guidance device, 22 ... Guidance signal generation part.

Claims (14)

It has two antennas per direction for monopulse angle measurement, receives and demodulates radar waves reflected by the target over multiple coherent integration periods, and outputs the demodulation results for each coherent integration period A radar receiver,
A target detection unit that detects a target from a demodulation result of each coherent integration period output from the radar receiver, and extracts a speed change in the plurality of coherent integration periods for the detected target;
A first data extraction unit for extracting a sum signal and a difference signal of the two antennas from the demodulation results of a plurality of coherent integration periods for the detected target;
An acceleration correction Fourier transform unit that performs a Fourier transform process to correct the plurality of sum signals and the difference signals extracted by the first data extraction unit based on the speed change extracted by the target detection unit;
A second data extraction unit for extracting one sum signal representative value and one difference signal representative value per target from the Fourier transform result;
A target angle detection apparatus comprising: a monopulse angle estimation unit that performs monopulse angle estimation from the sum signal representative value and the difference signal representative value.   The acceleration correction Fourier transform unit corrects the sum signal and the difference signal of the plurality of coherent integration periods based on the speed change, and Fourier transforms the corrected sum signal and difference signal. The target angle detection device according to 1. The acceleration correction Fourier transform unit performs fast Fourier transform on the sum signal and the difference signal corrected based on the speed change,
The second data extraction unit is configured to obtain a Fourier transform result of the sum signal of the bin number that gives a power maximum value of a spectrum having a larger power maximum value among the fast Fourier transform result spectra of the sum signal and the difference signal. 3. The target angle detection apparatus according to claim 2, wherein a complex value is used as the sum signal representative value, and a complex value of a Fourier transform result of the difference signal having the same bin number is used as the difference signal representative value.   2. The target angle detection device according to claim 1, wherein the acceleration correction Fourier transform unit performs discrete Fourier transform by changing a kernel interval during Fourier transform so as to correct the speed change. The acceleration correction Fourier transform unit sets a plurality of bins at a frequency interval equal to or less than an interval obtained by dividing the reciprocal of the coherent integration period by the number of the sum signal or the difference signal used for discrete Fourier transform, and the sum signal And each of the difference signals is discrete Fourier transformed in each bin to generate a spectrum,
The second data extraction unit includes a complex value of a Fourier transform result of the sum signal having a bin number that gives a power maximum value of a spectrum having a larger power maximum value among the spectrum of the sum signal and the spectrum of the difference signal. 5. The target angle detection apparatus according to claim 4, wherein the sum signal representative value and a complex value of a Fourier transform result of the difference signal having the same bin number are used as the difference signal representative value.   4. The acceleration correction Fourier transform unit performs Fourier transform in response to the speed change detected by the target detection unit and one or more speed changes in the vicinity thereof, respectively. Item 6. The target angle detection device according to Item 5.   The second data extraction unit has a speed corresponding to a spectrum having a maximum power value among a plurality of spectra of the sum signal and the difference signal obtained as a result of Fourier transform corresponding to a plurality of speed changes. 7. The target according to claim 6, wherein a complex value of a bin number in which the maximum power of the sum signal spectrum and the difference signal spectrum of the change value is maximum is used as the sum signal representative value and the difference signal representative value. Angle detection device.   The target angle detection apparatus according to claim 1, wherein the target detection unit detects a target locus from the demodulation results of the plurality of coherent integration periods.   9. The target angle detection apparatus according to claim 8, wherein the target detection unit detects a target by track-before-detect. The radar receiver is a pulse Doppler receiver,
The target angle detection apparatus according to claim 9, wherein the target detection unit detects a target from range information and Doppler frequency information in the demodulation result.   The said 1st data extraction part extracts the said sum signal and the said difference signal from the demodulation result of each coherent integration period corresponding to the locus | trajectory of the target which the said target detection part detected. The target angle detection device according to any one of claims 8 to 10.   9. The acceleration correction Fourier transform unit performs Fourier transform on a sum signal and a difference signal of a part of a plurality of coherent integration periods used by the target detection unit to detect a target locus. The target angle detection device according to any one of 11. Receive and demodulate radar waves reflected by the target over multiple coherent integration periods,
A target is detected from the demodulation result, and a speed change of the detected target in the plurality of coherent integration periods is extracted;
For the detected target, extract a sum signal and a difference signal of two antennas for use in monopulse angle measurement from the demodulation results of a plurality of coherent integration periods,
A plurality of the sum signal and the difference signal are Fourier transformed to correct the speed change,
Extract one sum signal representative value and difference signal representative value per target from the Fourier transform result,
A target angle detection method, wherein angle estimation by monopulse is performed from the sum signal representative value and the difference signal representative value. A target angle detection device according to any one of claims 1 to 12,
A guidance device comprising: a guidance signal generating unit that generates a guidance signal using at least an angle estimated by the target angle detection device with respect to a target detected by the target angle detection device.

Images