RU2693048C1 - Radar targets on the background of underlying surface selection method - Google Patents

Abstract

FIELD: radar ranging.

SUBSTANCE: invention relates to the radar ranging and can be used in radars, which provide for the scattering complete polarization matrix (PM) obtaining. Radar targets on the background of the underlying surface selection method comprises the target irradiation with two linearly orthogonally polarized waves, the reflected waves reception, their division into the vertical and horizontal components, forming the matrix based on polarization characteristics, wherein the target and the background irradiation is carried out by the linearly polarized waves, emitted in frequency band of not less than 20 MHz, the scattered waves division into the vertical and horizontal polarization components, at that, recording amplitudes of horizontally emitted polarization – horizontally received, of horizontally emitted polarization – vertically received, vertically emitted polarization – horizontally received, of vertically emitted polarization – vertically received, forming the complete PM for each i-th range strobe, which size is determined by the probing signal spectrum width, successively obtained scattering PM for each i-th range strobe are converted into the observation matrix A, which components are the signal amplitudes complex values, at that, the observation matrix each row corresponds to the scattered signal polarization components, and the lines number corresponds to the processed range resolution elements number, matrix dimension 4_*n, where n is the processed strobes number, normalizing the obtained matrix, forming the correlation matrix, for which calculating the eigen values and the eigen vectors, forming, thus, the new coordinate system, and then on the obtained new coordinate system axes projecting the normalized observation matrix A, as a result, obtaining the observations contributions matrix VK over the range, which is used for the target selection on the underlying surface background, at that, the target search and detection is first performed by signal level in the observations contributions matrix first column, and for the target against the underlying surface identification the rest columns signals are used.

EFFECT: increase in the radar targets detection reliability.

1 cl, 6 dwg, 5 tbl

Description

The invention relates to radar and can be used in radars, which provide a full polarization scattering matrix.

Modern methods for selecting radar targets are based on an analysis of the statistical characteristics of signals scattered by a target. At the same time, one of the informational characteristics is the polarization dispersion matrix (PMR) and the analysis of its statistical characteristics. The majority of studies performed both in Russia and abroad analyze the characteristics of the PMR for one element of the resolution

There is a method of selection of radar targets when controlling the movement of air and sea transport against the background of interfering reflections and interferences (see RF patent №2256194 of July 10, 2003, IPC G01S 13/04). The method of selection is that a radar target with known polarization parameters is irradiated with linear polarization signals with a rotating plane of polarization. In the signal reflected from the target, the second and fourth signals are separated relative to the frequency of rotation of the polarization plane of the original harmonic signal, whose parameters (amplitudes and phases) are the polarization parameters of the target. The decision on the presence or absence of a selectable goal is made on the basis of a decision rule, which is a linear combination of polarization parameters with coefficients determined by the statistical characteristics of polarization parameters with and without the goal. The decision decision threshold of the decision rule is determined by the maximum likelihood criterion and is assumed to be equal to one. Achievable technical result is an increase in the efficiency of selection of a target with known polarization parameters, i.e. for a given probability of false alarm, reducing to a minimum the probability of making a decision about the absence of a target, if present (the probability of missing a target).

There is a method of detection and selection of radar signals on the polarization basis (see RF patent №2476903 from 09.03.2011 IPC: G01S 13/04)

The invention consists in receiving two signal components that are orthogonal in polarization, converting analog signals into digital form, storing them in memory devices, interpolating digital signals, storing interpolated digital signals, and subsequently determining the ratio of amplitudes and phase differences of the selectable signal, calculating the total polarization parameter the received signal - the angle of ellipticity and the decision on the presence or absence of a selectable signal la in accordance with the Neumann-Pearson criterion.

The disadvantage of this method is that the detection of radar targets is performed according to the energy cumulative parameter (this does not take into account a priori probabilities of polarization of the reflected signal), which does not allow the detection and selection of reflected signals that are matched to polarization.

The technical task to implement the selection of radar targets on the background of the underlying surface is to develop an algorithm to separate the components of the polarization characteristics of the targets and the background based on a joint analysis of the scattering matrices across the range gates (resolution elements ΔR) in a given angular direction.

In the present invention is considered a set of consecutive elements of the resolution and, accordingly, PMR, belonging to each of them. Such a selection algorithm can be implemented using a high-resolution radar using a broadband signal, while the range resolution is at least 8 meters.

для каждого i-го строба дальности, размер которого определен шириной спектра зондирующего сигнала, затем последовательно полученные поляризационные матрицы рассеивания для каждого i-го строба дальности преобразуют в матрицу наблюдений А следующего видаTo solve the problem in the method of selection of radar targets on the background of the underlying surface, which consists in irradiating the target with two linearly orthogonally polarized waves, receiving reflected waves, dividing them into vertical and horizontal components, forming a matrix based on polarization characteristics, in which the target and background irradiation produce linearly polarized waves emitting in the range of the broadband frequency spectrum of at least 20 MHz, record the amplitudes u _GG - horizontally emit polarized - horizontally accepted, u _HS - horizontally radiated polarization - vertically received, u _HS - vertically radiated polarization - horizontally received, u _BB - vertically radiated polarization - vertically accepted, and the formation of a full polarization matrix

for each i-th range gate, the size of which is determined by the width of the spectrum of the probing signal, then successively obtained polarization scattering matrices for each i-th range gate are transformed into the following type of observation matrix A

видаwhere u is the complex amplitude of the signal, with each row of the matrix of observations corresponding to the polarization components of the scattered signal, and the number of rows corresponds to the number of processed elements of the range resolution, the matrix dimension is 4 _* n, where n is the number of processed gates of the range, the resulting matrix is normalized and then formed the correlation matrix for which the eigenvalues and eigenvectors are calculated, thus forming a new coordinate system, and then on the resulting axes of the new system to ordinate projecting normalized observation matrix A and normalization is performed as a result of deposits obtained from the observation matrix VK

kind of

which is used to select a target against the background of the underlying surface, while the search and detection of the target is primarily carried out by the signal level in the first column of the observation contribution matrix, and the signals from the remaining columns are used to recognize the target against the underlying surface.

The radar provides a full polarization matrix in each range gate in a given azimuthal direction, the size of the range gate does not exceed a certain predetermined value - resolution. It is assumed that the resolution is not worse than 8 meters (the width of the signal spectrum is at least 20 MHz). Generally speaking, the magnitude of the resolution in distance ΔR depends on the dimensions of the target L and the relation ΔR ≤ L. must be satisfied.

For each i-th strobe range receive the full polarization scattering matrix P _i

Successively obtained scattering matrices are expanded into a row and form the following observation matrix

There are four states in total. The first index corresponds to the number of the element of resolution in range, the entire elements of the range - M. In this case, the values of u _{i, j} can be both complex and valid. Further, a more general case is considered — the complex values of u _{i, j} . In this matrix, the second element index corresponds to the polarization state: 1 — horizontal polarization of the transmitter and horizontal polarization of the receiver (YY); 2 - horizontal polarization of the transmitter, vertical polarization of the receiver (GW); 3 - vertical polarization of the transmitter, vertical polarization of the receiver (BB); 4 - vertical polarization of the transmitter, horizontal polarization of the receiver (VG).

The resulting matrix of observations (2) is normalized. To do this, calculate the expectation of mo _j on the columns (polarizations) by the formula

and variance

The elements of the matrix of normalized values of UN are calculated by the formula

Then, in accordance with the algorithm of the principal component method, the correlation matrix is calculated using the formula

where the symbol * means complex conjugation, and T is the matrix transposition. As a result, we obtain a 4 × 4 matrix, which is Hermitian.

For this matrix, eigenvalues and, corresponding to them, eigenvectors are calculated.

As a result of the calculations, we have the vector of eigenvalues -SZ:

values of which are real and non-negative, and order them in descending order. For these eigenvalues (7), we calculate the eigenvectors of SVU and form the matrix of eigenvectors of the form

The resulting eigenvectors form a new coordinate system. We calculate the projections of the normalized matrix of observations UN on these new coordinate axes, which are commonly called the factors Fk. In matrix form, the operation is written as

Perform normalization

and form the matrix of contributions of observations VK

The resulting matrix (11) contains four columns, by the number of factors and M rows, corresponding to the distance gates.

In this case, in each column, the values are normalized to unity. That is, the sum of the elements of each column is equal to one.

The invention is illustrated in the drawing, which shows:

in fig. 1 is a view of the radar scene for one azimuth direction and the location of range gates,

in fig. 2 - block diagram of the algorithm,

in fig. 3 - model of the radar scene,

in fig. 4 - Values of ESR (effective area of dispersion) for the model

scenes that

in fig. 5 - eigenvalues of the correlation matrix,

figure 6 - contributions of the observations.

Target detection and its selection against the background of natural formations using the matrix of contributions (11) is performed as follows.

The signal level in the first factor (the values in the first column) (see Fig. 2) is determined by the total energy of the reflected signal, therefore, the search and detection of a target is primarily performed by the signal level in the first column. To recognize the target against the underlying surface, the signals from the remaining columns are used.

A regular target (consisting of a set of reflectors) responds in the first and second factors, that is, the values of the first and second columns for the corresponding range gate (matrix rows) exceed the detection thresholds. At the same time, the values for the third and fourth factors (in the third and fourth columns, respectively) do not exceed the threshold level.

Irregular targets that contain a variety of disordered elements, such as grass, shallow vegetation, and trees, are mainly grouped in the third and fourth factors. In addition, at high resolution (1 m), a mirror image of the target from the surface can be traced.

In the case when it is required to detect a target on the surface and not to classify the surface itself, the detection algorithm can be represented as

The detection thresholds for P1 and P2, apparently, should be selected according to the results of experiments or using an adaptive algorithm with a constant level of false alarms. When modeling and debugging the selection algorithm, the following values can be taken as a basis

This algorithm is illustrated with an example, where the implementation of the method for selecting a target on the model of a radar scene is shown (Fig. 1)

The object (self-propelled howitzer Palladin) is located on an uneven surface among bushes and wood).

The angular sector in azimuth is 0.3 degrees, the distance resolution of 1 m, a total of 40 range gates were simulated.

As a result of the simulation, for each range gate, full polarization matrices of the form were obtained.

where σ _{i, PI} - EPR in the i-th resolution element with the corresponding polarization of radiation (I) and reception (P), taking into account the reflections.

For the radar scene model, the calculation results have the form shown in Table 1. In the columns of the table, the values of the elements of the polarization scattering matrix obtained from the simulation results. The rows of the table correspond to the range gates. For real radar measurements, this table corresponds to sounding in one azimuthal direction over all distance gates during the accumulation period.

It is not difficult to see (Table 1 and Fig. 4) that the scattering matrices are not symmetric, the elements of the secondary diagonal are not equal to each other, since the reflections are taken into account in the simulation. The signal from the target and the natural formations is mainly focused on coordinated polarizations and there is no separation of the target and the natural formations.

For table 1 of the original data (matrix of observations), the correlation matrix has the form given in table 2.

The following eigenvalues are obtained for this correlation matrix.

In accordance with the distribution of dispersions obtained (Table 3) and values of eigenvalues (Fig. 5), one could limit oneself to two first eigenvalues or even one, since they explain almost 99% of the variance.

At the same time, we note that the eigenvectors and eigenvalues of the correlation matrix have a number of useful properties: 1) linearly independent, which allows them to be used as a basis; 2) the eigenvalues are non-negative and real; 3) the eigenvectors are orthogonal to each other and form an orthonormal set.

The sum of the eigenvalues of the correlation matrix is equal to the trace of this matrix, and their product is the determinant.

Consider the matrix of eigenvectors formed by the vector of eigenvalues. That is, in essence, the matrix of transition from the polarization basis to the factor basis.

It may be noted the following.

Factors 1 and 2 - you can attribute the physical meaning of the total energy

Factor 3 - the influence of the main polarizations

Factor 4 - the effect of cross-polarization.

The next step is to calculate the contribution of the observations (polarization matrix by range gates) to the relevant factors. In fact, the calculation of projections on new coordinate axes are factors.

The contributions of the observations in tabular form are shown in Table 5 and in FIG. 6

From tables 1 and 5 and the drawing it is clear that the target is concentrated on the factor coordinates 1 and 2, while background formations such as grass, shrubs and trees are concentrated in factors 3 and 4.

FIG. 6 that the goal is concentrated in the first and second factors, and natural formations in the 3rd and 4th factors.

Thus, the use of the described method of selection

radar targets against the underlying surface allows you to increase the accuracy of target recognition based on a joint analysis of polarization scattering matrices for range gates (resolution elements ΔR) in a given angular direction.

Claims (5)

The method of selecting radar targets against the underlying surface, consisting in irradiating a target with two linearly orthogonally polarized waves, receiving reflected waves, dividing them into vertical and horizontal components, forming a matrix based on polarization characteristics, characterized in that the target and background irradiation produce linearly polarized waves emitting in the range of a broadband frequency spectrum of at least 20 MHz, they separate the scattered waves into vertical and horizontal polarization leaving, recorded amplitude

- horizontally radiated polarization - horizontally accepted,

- horizontally radiated polarization - vertically adopted,

- vertically radiated polarization - horizontally adopted,

- vertically radiated polarization - vertically adopted, form a complete polarization matrix

for each i-th range gate, the size of which is determined by the width of the spectrum of the irradiating signal, then successively obtained polarization scattering matrices for each i-th gate of the range are converted into an observation matrix A of the following form:

, where u is the complex amplitude of the signal, while the rows of the matrix of observations correspond to the polarization components of the scattered signal, and the number of rows corresponds to the number of processed elements of the range resolution, the matrix dimension is 4 * n, where n is the number of processed distance gates, the resulting matrix is normalized and a correlation matrix is formed , for which eigenvalues and eigenvectors are calculated, thus forming a new coordinate system, and then on the resulting axes of the new coordinate system simulate the normalized matrix of observations A and perform the normalization; as a result, a matrix of contributions of the observations VK in the range from (νk _1.1 ... νk _1.4 ) to (νk _M.1 ... νk _M.4 ) is _obtained

which is used to select a target against the background of the underlying surface, while the search and detection of the target is primarily carried out by the signal level in the first column of the observation contribution matrix, and the signals from the remaining columns are used to recognize the target against the underlying surface.

Images