RU2593146C1 - Method for adaptive detection of signals of moving targets on background of multicomponent passive interference - Google Patents

Abstract

FIELD: radar ranging and radio navigation.

SUBSTANCE: invention relates to digital processing of radar signals. Said technical result is achieved by that for multichannel Doppler filtration and multichannel coherent accumulation in form of Fourier transform, weight coefficients are calculated in real time based on estimates of coefficients of autoregression by averaging over several elements of range. After that envelopes of signals at output of each channel are calculated, which is standardised and combined with extraction of maximum value. Detection threshold is compared at output in each range element with maximum from several maxima of envelope signals obtained from processing each bundle of pulses with different repetition frequencies or carrier frequencies, variable from packet to packet.

EFFECT: high efficiency of detecting moving targets on background of multicomponent passive interference caused by a set of reflections from local objects, clouds, hydrometeors, dipole interference.

3 cl, 2 dwg

Description

The proposed method relates to radio engineering, in particular to digital processing of radar signals. In practice, radar protection from multicomponent passive interference signals caused by a combination of reflections from local objects, clouds, hydrometeors, dipole interference, various methods for detecting useful targets are used.

A known method of detecting signals against a background of passive interference, based on preliminary adaptive filtering of passive interference using the appropriate selection of filter weights [1] and then comparing the envelope of the output signal with a threshold. The main disadvantage of this method is its low efficiency, due to the use of fixed weights for filtering signals having a Doppler shift of the spectrum. In addition, the absence of signal normalization before comparison with the threshold leads to a lack of stabilization of false alarms.

A known method of detecting signals against a background of passive interference [2], which includes multi-channel coherent weight accumulation, envelope separation after multi-channel coherent weight accumulation, adaptive normalization of signal envelopes in each channel and their combination, thereby forming an output signal compared with a threshold detection.

The disadvantage of this method is the reduced efficiency of protection against passive interference compared to optimal processing due to the fact that the formation of the type of frequency characteristics during coherent weight accumulation is carried out without taking into account parameters such as the width of the spectrum of fluctuations of passive interference and Doppler shift of its frequency. The use of the same weight function in all channels to reduce the level of the side lobes of the frequency characteristics of the channels provides a symmetric and uniform suppression of all side lobes of the filters, while the optimal processing maximizes the signal-to-noise ratio due to different correction of the frequency characteristics of the filters only in the spectral zone fluctuations in passive interference signals.

The closest in technical essence to the proposed method is a US patent [3], in which the processing is constructed as follows. At the input, the reflected signals represented by their digital quadrature components are subjected to multichannel Doppler filtering and multichannel coherent accumulation in the DFT processor, at the output of which envelopes are calculated, which are normalized, combined, and fed to the detection threshold, and weighting coefficients for multichannel Doppler filtering are preliminarily calculated using an approximate formula for a previously selected shape of the spectrum of passive interference fluctuations. In this method, the correction of the frequency characteristics during multichannel Doppler filtering is performed more optimally taking into account passive interference, which is close to reality, having, for example, a Gaussian spectrum. Why pre-weighting factors that provide optimization of frequency characteristics are calculated by the formula (see patent description)

,

,

where W is the vector of weights for multichannel Doppler filtering,

R ^-1 - inverse correlation matrix for a predefined model of passive interference and stored in read-only memory,

S is the vector of the useful signal, which for the target with an unknown speed is formed in the form of the Fourier coefficient

S = (1, e ^{j2πn / N} , e ^{j4πn / N} , ..., e ^{j2π (N-1) / N} ),

where n is the number of the Doppler channel, N is the number of pulses in the packet.

In the method taken as a prototype, the calculation of weighting coefficients due to the impossibility of real-time estimation of a multidimensional complex correlation matrix and making it inverted, a simplified approach was used based on a model of passive interference, for example, with a Gaussian shape of the spectrum and calculation of weighting coefficients without an estimate and inverting the entire correlation matrix in real time. Instead, only the modulus of the inter-period correlation coefficient and its argument, the inter-period Doppler interference phase shift, are estimated in real time. And based on them, pre-calculated weights for the stationary passive interference are selected from the memory device, which are adjusted taking into account the estimates of the inter-period Doppler phase incursion. Unfortunately, it is impossible to foresee the whole variety of spectrum shapes, especially if the passive interference is multicomponent, i.e. It has a multimode spectrum of reflected signals simultaneously from local objects, hydrometeors and dipole reflectors. Hence the low efficiency of this method

To increase the efficiency of radar protection from passive interference with multimode spectra, it is proposed to calculate the Doppler filter weights from the reflected signals from a multicomponent passive interference using an approximation of real interference by the autoregression process [4]. This will also allow not to evaluate the correlation matrix and its inversion, but to evaluate autoregression coefficients in real time. We use in our case a different representation of formula (1), avoiding the estimation and inversion of the complex correlation matrix. The inverse correlation matrix for the autoregression process can be represented as a decomposition into diagonal D and upper and lower triangular matrices of autoregression coefficients A:

In accordance with (2) we obtain:

where Z _in , Z _{out are} complex signals represented by their quadrature components at the input and output of the Doppler filter, A is the complex vector of autoregression coefficients. The left side of formula (3) in terms of the prototype patent can be considered as multichannel Doppler filtering, and the right as a coherent multichannel accumulation in the DFT processor. And the diagonal terms D are the result of normalizing the signals in each filter channel.

We transform (3) so that we compatible Doppler filtering with coherent accumulation of target signals with an unknown speed

where n is the number of the Doppler channel, S (n) are the Fourier coefficients

S (n) = (1, e ^{j2πn / N} , e ^{j4πn / N} , ..., e ^{j2π (N-1) / N} ),

N is the number of pulses in a packet.

Thus, in the proposed method, at an unknown target speed, adaptive processing is implemented using multichannel Doppler filtering, for each channel in which its weight coefficients are used, taking into account both autoregression coefficient estimates and Fourier expansion coefficients, allowing filtering passive signals, unlike the prototype interference with multimode spectra.

Thus, in the known method of adaptive target detection against the background of multicomponent passive interference, including multichannel Doppler filtering, with multichannel coherent accumulation in the form of a discrete Fourier transform, the result of which is subjected to the calculation of envelopes that are normalized and combined with the allocation of the maximum value, significant differences are introduced in that, for the multi-channel Doppler filtering operation, the weight coefficients are calculated in real time vector multiplication of the estimated autoregression coefficients, and the Fourier transform coefficients.

This is the most important advantage and a distinctive feature of the proposed method, since in this case it is possible to obtain real-time estimates of autoregressive coefficients using a computationally efficient algorithm without preliminary estimation of the interference correlation matrix and its inversion, using directly the observation sample. At the same time, using a limited number of autoregression coefficients, it is possible for a wide class of multicomponent passive interference having multimode spectra to provide effective target detection. The most preferred of the algorithms for estimating autoregression coefficients from the point of view of efficiency is the Berg algorithm [5].

Another distinctive feature of the proposed method is the processing of several bursts of pulses at different repetition frequencies or at different carrier frequencies, which allows to improve the speed characteristics of the signal detector on the target as a whole. For this, when processing each burst from M, the signal after selecting the maximum of each processed burst of pulses is remembered and, at the end of the last processed burst, the maximum from all the obtained maxima is selected. It is compared with the detection threshold.

MAX {MAX ₁ | Z _1BX W ₁ (n) |, MAX ₂ | Z _2BX W ₂ (n) | ... MAX _M | Z _MBB W _M (n) |} ≥U _POR

In order to eliminate the suppression of the useful signal during filtering, it is proposed to average the estimates of autoregression coefficients over several range elements.

, вторая мода $$\raisebox{1ex}{$\Delta F_2$}\!\left/ \!\raisebox{-1ex}{$F_{П}=\mathrm{0,25}$}\right.$$

. Каждая мода имела превышение над шумом 20 дБ с относительным доплеровским смещением $$\raisebox{1ex}{$F_{Д1}$}\!\left/ \!\raisebox{-1ex}{$F_{П}=\mathrm{0,35}$}\right.$$

и $$\raisebox{1ex}{$F_{Д2}$}\!\left/ \!\raisebox{-1ex}{$F_{П}=\mathrm{0,5}$}\right.$$

. Скорость цели задавалась 500 м/с. Число обрабатываемых импульсов 12 (две пачки по 6 импульсов на двух частотах повторения 400 Гц и 440 Гц.) Порядок авторегрессии был задан 5. Оценка коэффициентов авторегрессии производилась по программе из МАТЛАБ [6] для первой SF11 и второй SF22 пачек:The gain in the effectiveness of the proposed method compared to the prototype was evaluated in a threshold signal for the probability of correct detection of 0.5 and the probability of false alarm 10 ^-1 -10 ^-2 . The calculation was performed by the method of statistical modeling in MATLAB. The passive interference was specified as extended in range, two-component (with a two-mode spectrum), the first mode had a relative width of the fluctuation spectrum at a level of -20 dB $$\raisebox{1ex}{$\Delta F_{\mathrm{one}}$}\!\left/ \!\raisebox{-1ex}{$F_P=0.15$}\right.$$

second fashion $$\raisebox{1ex}{$\Delta F_2$}\!\left/ \!\raisebox{-1ex}{$F_P=0.25$}\right.$$

. Each mode had an excess of 20 dB over noise with a relative Doppler shift $$\raisebox{1ex}{$F_{D\mathrm{one}}$}\!\left/ \!\raisebox{-1ex}{$F_P=0.35$}\right.$$

and $$\raisebox{1ex}{$F_{D2}$}\!\left/ \!\raisebox{-1ex}{$F_P=0.5$}\right.$$

. The target speed was set at 500 m / s. The number of processed pulses is 12 (two bursts of 6 pulses at two repetition frequencies of 400 Hz and 440 Hz.) The autoregression order was set to 5. The autoregression coefficients were estimated using the program from MATLAB [6] for the first SF11 and second SF22 bursts:

Weighting coefficients were generated by vector multiplication of estimates of autoregression coefficients by Fourier coefficients

In FIG. Figure 1 shows the frequency response of one of the channels of a Doppler filter using estimates of autoregressive coefficients. For the method taken by the prototype, the module of the inter-period correlation coefficient and the inter-period Doppler phase incursion were estimated using the MATLAB program

These estimates, assuming a Gaussian shape of the interference spectrum, were used to reproduce, according to the module estimates, the inter-period correlation coefficient of the interference correlation matrices without Doppler shift for the first and second bursts, which are then accessed using the inv MATLAB program. Below is given as an example the correlation matrix reconstructed for the first burst according to the estimate of the module of the inter-period correlation coefficient equal to 0.7452:

And the corresponding inverse matrix

After the rotation of the vector of the input sample of observations by an angle in accordance with the obtained estimate of the argument of the inter-period correlation coefficient, multichannel Doppler filtering is performed with weights, using reproduced inverse matrices RE1 for the first packet and RE2 for the second packet:

Since the description of the prototype method considers the processing of only one pack, therefore, at first, the comparison of the proposed method with the prototype was performed when processing one pack, i.e. when the selected envelopes at the output of the Doppler channels after averaging were combined with the selection of the maximum value using the max MATLAB function, which was fed to the detection threshold

Then, the same comparison of the two methods was made when processing two packs with different two repetition frequencies with selection, in accordance with the formula of the proposed method, the maximum of two maximums when processing each pack.

The calculations showed that the gain in the threshold signal of the proposed method compared to the method taken as a prototype when processing one pack, and for two packs, is 5-10 dB. To explain the obtained higher detection efficiency is quite simple if we compare the correlation functions of each of the two components of the passive interference with the correlation function obtained by estimating the resulting module of the inter-period correlation coefficient of the two-component passive interference (see Fig. 2). In addition, an estimate of the inter-period Doppler phase incursion (the argument of the inter-period correlation coefficient) is also erroneously formed by two interference components shifted by the Doppler. Hence, the “optimal” weighting coefficients in the prototype do not correspond to the real interference environment, which cannot be foreseen in advance, even having saved a huge number of ready-made weighting factors in the storage device, which leads to a decrease in the prototype’s efficiency of detecting useful targets against the background of multicomponent passive interference.

Information sources

1. Bakulev P.A., Stepin V.M. Methods and devices for moving targets selection. - M., "Radio and Communications", 1986.

2. US Patent No. 3831174 G01s 9/06, 1974.

3. US patent No. 4742353 G01s 13/6, 1988.

4. Marple S.L. "Digital Spectral Analysis", Englewood Cliffs, NJ: Prentice Hall, 1987, Chapter 7.

5. Bartenev V.G., Kutepov V.E. A comparative analysis of two methods of forming weight coefficients in an adaptive system for moving target selection // Digital signal processing. 2014. No. 2, S. 58-60.

6. Potemkin V.G. "MATLAB Handbook" Analysis and data processing. http://matlab.exponenta.ru/ml/book2/chapter8/.

Claims (3)

1. The method of adaptive detection of moving targets against the background of multicomponent passive interference, including multichannel Doppler filtering with multichannel coherent accumulation in the form of a Fourier transform, the result of which is calculated envelopes, which are normalized and combined with the allocation of the maximum value, characterized in that for the multichannel operation Doppler filtering and multichannel coherent accumulation in the form of a Fourier transform, the weight coefficients are calculated in real m-time vector by multiplying ratings autoregressive coefficients and Fourier transform coefficients. 2. The method of adaptive detection of moving targets against the background of multicomponent passive interference, including multichannel Doppler filtering with multichannel coherent accumulation in the form of the Fourier transform, the result of which is calculated envelopes, which are normalized and combined with the allocation of the maximum value, characterized in that with a detection threshold compares at the output in each element of the range a maximum of several maximums of the envelopes of the signals obtained by processing each packet e quidistant pulses with different repetition frequencies or carrier frequencies that vary from pack to pack. 3. The method according to p. 1, characterized in that the formation of estimates of the coefficients of autoregression is performed by averaging these estimates of the coefficients of autoregression over several elements of the range.

Images