RU2774733C1 - Method for classifying mobile objects of ground equipment using the peculiarities of their adhesion to the soil - Google Patents

Abstract

FIELD: seismic studies.

SUBSTANCE: invention relates to the field of seismic exploration and can be used in security and reconnaissance-alarm complexes and systems for processing seismic signals generated by objects of ground equipment (OGE) in the surface layer of soil. A method for classifying moving objects of ground equipment by a seismic signal is proposed, including registration of a seismic signal excited by a moving object of ground equipment, filtering it in a selected frequency band, amplification and sampling. The method also includes the selection of the operating frequency range ΔF, characteristic of seismic vibrations of the soil, excited by the OGE of the “wheel” or “caterpillar”, as well as the division of the total frequency range ΔF into two bands Δf ₁ and Δf ₂ so that in each of these bands, most of the energy of seismic waves from “their” OGE class concentrates. The value of the signal energy ratio in these frequency bands, reduced to the value of the signal-to-noise ratio in the first frequency band, relative to some statistically reasonable level is a classification feature of the detected OGE.

EFFECT: adaptation of the OGE classification process to changes in the level of external noise, interference or weather and climatic conditions, which increases the robustness of the method, the probability of correct classification of OGE and reduces the likelihood of confusing OGE classes.

1 cl, 10 dwg

Description

The invention relates to the field of seismic reconnaissance and can be used for target designation of weapons, in control devices for the detonation of engineering ammunition, as well as for monitoring terrain and controlling unauthorized access to protected objects.

There are various methods for classifying moving ground objects based on the results of the analysis of seismic signals excited by them in the surface layer of the soil in the process of movement. For example, in [1], [2] and [3] the operation of devices is described, which characterizes the methods for classifying moving objects in a controlled area, which differ in the features of searching in a seismic signal for various feature parameters to be used in classifying objects.

In [1], for this, an autoregressive model is used, taking into account the fact that random processes can be described as an autoregression of a certain order, characterizing the degree of influence of previous signal samples on the current sample.

According to the accepted implementation of the seismic signal, the regression coefficients are estimated and compared with those stored in the database of averaged models, which makes it possible to calculate the conditional probability of the object belonging to a certain class based on the results of the analysis of the received seismic signal. The decision on the class of the object is made according to the criterion of the maximum conditional probability.

The disadvantages of this method include the following:

- the need to store averaged models in the data bank

regression coefficients and information about the laws of their distribution;

- the number of classification channels should be equal to the number of object classes;

- cumbersomeness of the computational algorithm;

- regression coefficients refer to parameters that depend on the type and condition of the soil, season, interference environment and other factors, therefore, when the properties of the soil change in the spectra of seismic signals, there will be a shift in the frequency components to the region of high or low frequencies, that is, reliable recognition of objects by analysis of such signals is possible when using the results of the analysis of the dynamics of the behavior of the seismic signal on a certain specified time interval.

In [2], the degree of proximity of seismic signals, the sources of which are the classes of moving objects to be recognized, to the reference seismic signals is estimated.

After the detection of the object, the seismic signal is normalized in amplitude and discretized taking into account the Kotelnikov theorem. Then it is subjected to a Fourier transform aimed at highlighting the classification features, which are the values of the average frequency components in the frequency ranges characteristic for determining classes of objects.

Frequency features are required to calculate the measure of proximity to seismic signal standards, the sources of which are object classes to be recognized, which are stored in the data bank. The values of the sums of the basis functions necessary to calculate the values of frequency features in real time are also stored there.

After the proximity measure is calculated, the object class is selected: according to the measure minimum criterion, two classes are selected with the preservation of information about the degree of proximity. Further, based on the analysis of only the frequency features of the seismic signal, using information about the frequency of occurrence of pairs of classes, taking into account the degree of proximity to the standard, a decision is made to recognize the class of the object. For this, information is used, which is statistical data on the frequency of occurrence of such pairs in the form of a matrix, which is also stored in the data bank.

The main disadvantages of the method are:

- the need to store in the data bank the values of the sums of basic functions and matrices containing statistical data on the frequency of occurrence of classes, taking into account the degree of proximity to the standard;

- with an increase in the number of classes of recognition objects, in order to increase the probability of correct classification, it is necessary to increase the number of features characteristic of the selected classes;

- cumbersomeness of the computational algorithm;

- frequency features are dependent on the type and condition of the surface layer of the soil, which is determined by both interference factors and weather and climatic conditions, therefore, when the state of the surface layer of the soil changes in the spectra of seismic signals, there will be a shift in the frequency components to the region of high or low frequencies, which is negative will affect the frequency features, and this will reduce the likelihood of correct classification of objects.

In [3], the method is based on the analysis of the dynamics of changes in the seismic signal over time. The presence of patterns in the behavior of the seismic signal (the rise and fall times of the fronts, the time the seismic signal stays in extrema, etc.). This makes it possible to conclude that the seismic signal will pass certain specified conditional thresholds in a certain sequence. Moreover, for different sources, the sequence of passage will vary significantly.

With a limited amount of information about the behavior of signals from all sources, an average image of a seismic signal (standard) is constructed based on the results of an analysis of several dozen realizations of seismic signals excited in the surface layer of the soil during the movement of any class of ground equipment. All other seismic signals that, as a result of the analysis, will not fit the specified rule with the specified accuracy value, are considered unrecognized or interference. The generated average signal image during operation is stored in the bank as a set of threshold voltage values and values of time intervals for passing these thresholds. The number of both is selected based on the requirements for the accuracy of classifying objects in the analysis of seismic signals: the greater the number of specified thresholds, the more accurately the features of the object are tracked in the analysis of the seismic signal.

After registration, the seismic signal is subjected to pre-filtering, amplified in the characteristic frequency bands, and it is discretized in accordance with the Kotelnikov theorem.

Next, the discretized seismic signal is analyzed in accordance with the algorithm stored in the data bank. It controls the fulfillment of a number of conditions:

- the thresholds that determine the required value of the signal amplitude must be exceeded;

- a number of temporal criteria of the signal impact (period, duration, pauses between pulses) should differ little from the reference ones.

When the sequence of time intervals for passing the thresholds coincides with the reference intervals stored in the data bank, a logical signal is generated about the recognition of a given object class.

If it is necessary to classify several types of objects, it is necessary to have the same number of classification channels, which will differ in that their banks store other values of the response thresholds and other time intervals for their passage.

The main disadvantages include the following:

- the need to store a set of reference values of threshold stresses and values of time intervals for passing these thresholds for comparison with the temporal structure of the received seismic signal;

- signal processing only in the time domain can lead to the fact that the impact of external factors (interference, weather and climatic phenomena, etc.) will affect the redistribution of the energy of the frequency components, which will distort the temporal structure of the seismic signal and lead to a decrease in the probability of correct object classification;

- the ability to implement detection based on seismic signals received at high seismic levels, which, when the source of seismic vibrations is removed, can reduce the values of useful seismic signals to the level of seismic outliers and, after detection, lead to false classification.

The closest analogue in technical essence to the present invention is the method described in [4], taken as a prototype.

In the prototype method, an option is proposed to reduce the probability of false classification of moving ground objects by introducing a detection threshold with a level value depending on the intensity of external interference.

As in [3], the prototype method is based on the dynamics of changes in the seismic signal over time. The difference is the experimental establishment of the distribution law for the controlled classification feature and its parameters in order to establish the degree of influence of the seismic noise level on the quality of object detection based on the results of the analysis of incoming seismic signals. As such a feature, the maximum values of pulse signals are used.

Preliminarily, based on the results of statistical processing of records of seismic background signals in the amount of 1 million 254 thousand samples at the output of a band-pass filter with the selected parameters, a hypothesis was put forward that the distribution of the envelope maxima at the output of such a filter obeys the law of distribution of maxima described by the expression

. (1)

. (one)

In this distribution, the parameters λ and μ are related to the mathematical expectation of the process and the standard deviation by the following dependencies

или

or

. (2)

. (2)

In this case, the ratio for the probability of false detection or the probability of exceeding the level V ₀ under the influence of the seismic background is found using the distribution function (1)

(3)

(3)

. Вероятность того, что за всё установленное время функционирования реализующего устройства Т _а произойдет ложное обнаружение на любом из интервалов nТ ₀, число которых равно N =T _a/nТ ₀ будетIf we control the simultaneous excess of the threshold level, for example, in two adjacent time windows during the time nТ ₀ ( n = 2), then the probability of independent events is equal to

. The probability that for the entire set operating time of the implementing device T _a a false detection will occur at any of the intervals nT ₀ , the number of which is N = T _a / nT ₀ will be

. (4)

. (four)

The prototype method is as follows.

The generated average image of the seismic signal during operation is stored in the data bank in the form of a set of threshold stress values and values of the time intervals for passing these thresholds, the number of both is selected based on the requirements for the classification accuracy of moving objects in the analysis of seismic signals.

In the absence of mobile objects of ground equipment.

1. The seismophone is filtered in the frequency band (16-32) Hz.

с использованием скользящего временного окна длительностью T ₀=0,6 с. Здесь p = 2, 3, …, N-1. N = T ₀·F _стр, где F _стр – частота стробирования. 2. The maximum of the envelope x _m is selected from the set of extreme values of the seismophone

using a sliding time window with a duration T ₀ =0.6 s. Here p = 2, 3, …, N -1. N = T ₀ · F _str , where F _str is the sampling frequency.

и среднее квадратическое отклонение σ _mn максимумов огибающей сейсмофона _3. The mathematical expectation

and the standard deviation σ _mn of the maxima of the seismophone envelope

.

.

4. The values of the parameters of the law of distribution of maxima are estimated

(5)

(5)

and the threshold level V ₀ with arguments (5), which provides the required probability of false detections P _l1 for the entire set time of operation of the implementing device T _a , is found using (3)

, (6)

, (6)

where the probability of false detections during a single analysis of the seismophone in two adjacent time windows in accordance with (4) will be equal to

. (7)

. (7)

Here, the value of P _l,ass is determined by the requirements for the implementing device in operating conditions.

It is shown that when the Kolmogorov criterion is fulfilled, the distribution of the envelope maxima obeys the distribution of maxima with numerical values of the parameters

μ = 1.196 10 ^-3 V, λ = 2.259 10 ^-4 V (8)

at values

и

.

and

.

In the presence of moving objects of ground equipment.

1. The received seismic signal is subjected to pre-filtering and amplification in characteristic frequency bands and, in accordance with the Kotelnikov theorem, it is discretized.

2. The discretized seismic signal is analyzed in accordance with a certain algorithm. In this case, just as in [3], the fulfillment of a number of the same conditions is controlled:

- exceeding the thresholds that determine the required value of the signal amplitude;

- sequence of temporal criteria of signal impact (period, duration, pauses between pulses), which should differ little from the reference sequence.

3. If the sequence of time intervals for passing the thresholds coincides with the reference intervals stored in the data bank, a logical signal is also generated about the recognition of a given object class.

If it is necessary to classify several types of objects, it is necessary to have the same number of classification channels, which will differ in that their data banks will store other values of the response thresholds and other time intervals for their passage.

In FIG. 1a shows dependence (3) using (6) and (8), and in Fig. 1b - dependence (6) on a logarithmic scale using (7) and (8) for T _a = 30 days and P _l,set = 10 ^-6 .

From the analysis of Fig. It follows from Table 1 that, in contrast to the results presented in [3], in the case of a decrease in the values of useful seismic signals to the level of seismophone emissions, by changing the value of the detection threshold V ₀ in the desired direction, it is possible to minimize the probability of false object classification.

The prototype method is characterized by some of the disadvantages, namely:

- the need to store a set of reference values of threshold stresses and values of time intervals for passing these thresholds for comparison with the temporal structure of the received seismic signal;

- signal processing only in the time domain can lead to the fact that the impact of external factors (interference, weather and climatic phenomena, etc.) will affect the redistribution of the energy of the frequency components, which will distort the temporal structure of the seismic signal and lead to a decrease in the probability of correct object classification.

The task of the proposed technical solution is to adapt the ONT classification process to changes in the level of external noise, interference or weather and climatic conditions, which increases the robustness of the method, the probability of correct classification of ground equipment objects (ONT) and reduces the likelihood of confusing ONT classes.

To solve the problem in a method for classifying moving objects of ground equipment by a seismic signal, including the registration of a seismic signal excited by a moving object of ground equipment, its filtering in the selected frequency band, amplification and sampling, according to the invention , in the absence of an object, the selected operating frequency range is divided into two bands - low-frequency and high-frequency according to the condition of equality in them of the spectral densities of the seismophone, periodic estimation of the average value of the seismophone energy in the selected band during the observation of the seismophone, as well as the value of the coefficient of equalization of the energies of the seismophone in these bands, as the reciprocal of the energy averaged over the observation time interval low-frequency band seismophone to the high-frequency band seismophone energy averaged over the same time interval; in the presence of a mobile object of ground equipment for the duration of the current time window, the energy value of the mixture of the useful seismic signal and the seismic phone in the low-frequency band is divided by the energy value of the mixture of the useful seismic signal and the seismic phone in the high-frequency band, the obtained quotient is multiplied by the last estimate of the leveling coefficient value and the current values are compared result with two thresholds - upper and lower; if these current values cross the lower or upper thresholds from top to bottom or from bottom to top, the fact of detection of an object of ground equipment is recorded, in the case when, after crossing these current values of the lower or upper thresholds, they remain between these thresholds, a variant of missing a useful seismic signal will be implemented - non-detection object, when all current values are above the upper threshold, the result is ignored; when a moving object is detected, the ratio of the current values of the energy of the mixture of the useful seismic signal and the seismophone in the low-frequency band to the average energy of the seismophone calculated during the last observation period in the absence of moving objects is calculated, and the result of this ratio is divided by the result of the ratio of the energy of the mixture of the useful seismic signal and the seismophone in the low-frequency band band to the same energy in the high frequency band; if the sequence of the minimum values of the result obtained, grouped in a small neighborhood around the traverse point, is below the statistically justified value of the classification threshold, then the object is classified as a "caterpillar", if below this threshold, then it is classified as a "wheel".

The seismic signals generated by ground equipment objects in the surface layer of the soil were considered during their passage at various traverse distances from a three-component seismic receiver; the spectra are analyzed when objects move in the vicinity of the traverse points. Based on the accumulated experience in this area, it was concluded that the rational width of the total band of the studied spectra can be limited to (30-35) Hz. At the same time, for objects of ground equipment of the “wheel” class, the most statistically stable component in this band is a component with a frequency f _CT = 15 Hz, while for objects of ground equipment of the “caterpillar” class, such a component is characterized by a frequency f _GT = 25 Hz. This is explained by the fact that during the movement of wheeled objects, vibrations of their bodies on suspensions (springs, etc.) through the adhesion of the wheels of the object to the ground generate a spectrum of frequencies in it, grouped around the frequency f _CT . During the movement of caterpillar objects, the impacts of numerous rollers on the part of the caterpillars that adhere to the ground generate a spectrum of frequencies in it, grouped around the frequency f _ГТ . In addition, it is known that the low-frequency part of the spectrum (approximately from 0 to 10 Hz) is initiated mainly by the seismophone, which has a variety of sources, such as echoes of distant volcanic activity and distant earthquakes; near wind impacts on the vegetation cover of the soil, etc. In this regard, when searching for informative parameters of the studied seismic signals, their spectra were considered within the limits of (10-32) Hz. Therefore, taking into account the Nyquist theorem, the sampling rate was chosen when digitizing analog signals equal to f _d = 64 Hz.

Based on the foregoing, the entire frequency band Δ f \u003d f ₁ - f ₂ \u003d 32-10 \u003d 22 Hz is divided into two such parts

Δf _one =f _x -f _one and Δf ₂ =f ₂-f _{x +1} = Δf - Δf _one,

for which the ratio of the seismophone energies contained in them, obtained by its accumulation over a period of observation of a certain duration, would be close to unity. Here f _x is the frequency dividing the band Δ f in the required ratio. Such an equilibrium state of noise energies in these bands makes it possible to determine the values of the energies of seismic signals excited by various classes of ground equipment objects in the surface soil layer, which are close to true. If this balance is disturbed, then it is necessary to periodically compensate for the negative effect of the redistribution of frequency components in reasonably selected low-frequency and high-frequency components of the common frequency band.

The proposed method includes the following operations.

In the absence or presence of ground equipment.

, либо смеси полезного сейсмосигнала и сейсмофона

, регистрируемых в аналоговом виде в полосе Δf, с частотой дискретизации f _d и получить три соответствующие дискретизированные компоненты

или соответственно

с количеством отсчётов в каждой, зависящим от величины времени наблюдения: i = 0, 1, ….1. Discretize three components of either natural seismic background

, or a mixture of a useful seismic signal and a seismophone

recorded in analog form in the band Δ f , with sampling frequency f _d and obtain three corresponding sampled components

or respectively

with the number of readings in each, depending on the value of the observation time: i = 0, 1, ….

2. Enter the number of frequency samples N _f = 32: n = 0, 1,… ,

N _f -1 - current count numbers,

introduce a sequence of sampling periods (sliding time windows on which N _p = 64 time readings fit), the number of which is also determined by the value of the observation time:

k = 0, 1, … - numbers of current positions of time windows

(9)

(9)

where h _k = kN _p .

Therefore, the values (9) are matrices, where each row is a time window at the k -th current position, containing the samples included in it from 0 to N _p -1.

In the absence of ground facilities.

1. Determine the rational value of the dividing frequency f _x .

1.1. At the selected duration of the seismophone observation time interval, containing N _u current positions of the time window, using the Fourier transform, find the current spectral densities of the three components of the natural seismic background in the band Δ f , determine the total spectral density of the seismophone in this band

. (10)

. (ten)

Here

(11)

(eleven)

- мнимая единица.current spectral densities x -, y - and z -components of the seismophone at the k -th position of the time window;

is the imaginary unit.

сейсмофонов, записанных в разные дни для произвольных позиций временного окна в выбранном диапазоне частот.As an example, in FIG. 2a and fig. 2b in this frequency band shows the moduli of the total spectral densities

seismophones recorded on different days for arbitrary positions of the time window in the selected frequency range.

: f _x ≈ 19 Гц. Это автоматически приводит к приближённому выполнению равенства энергий сейсмофона в этих полосах. ЗдесьThe analysis shown in Fig. 2, as well as many similar ones, showed that based on the spectral density of the seismophone, it is possible to determine with sufficient accuracy the desired average value of the dividing frequency, which provides the value of the ratio of the averaged total spectral densities of the low-frequency and high-frequency bands of the total spectrum

: fx _≈ 19 Hz. This automatically leads to an approximate fulfillment of the equality of the seismophone energies in these bands. Here

are the spectral densities of the seismophone averaged over all N _u current positions of the time window in the low-frequency and high-frequency bands. Brackets < * > mean the operation of averaging over the duration of the time window (64 samples).

1.2. In the found frequency bands Δ f ₁ = 19-10=9 Hz and

, и

, Δ f ₂ \u003d 32-20 \u003d 12 Hz find the current values of the total energies of the seismophone

, and

,

where

(12)

(12)

energy seismophone component in the respective frequency bands.

1.3. Provide filtering of these energies with a complementary filter with gains K ₁ = 0.9 and K ₂ = 0.1:

. (13)

. (13)

1.4. Find the ratio of the current values of the filtered seismophone energies in the general case unequal to unity

.

.

1.5. Calculate the average value over all current positions of the time window

. (14)

. (fourteen)

1.6. Find the equalizing factor as the reciprocal of (14)

.

.

1.7. Compensate for the influence of external factors on the distribution of the seismophone energy over the frequency components of the bands (restore the energy balance relative to the equilibrium threshold equal to one)

. (15)

. (fifteen)

1.8. Periodically evaluate the value of relation (14), find the current value of the equalizing factor and use it to restore the balance of the seismophone energies in the frequency bands Δ f ₁ and Δ f ₂ if necessary.

When exposed to seismic signals from ground equipment.

1. By analogy with (14), taking into account the use of the equalization coefficient found at the last estimated observation time interval, determine the ratio of the energy of the mixture of the useful seismic signal and the seismophone in the low-frequency band to the same energy in the high-frequency band

. (16)

. (16)

2. Track the current values of the adjusted ratio (16) and compare them with two thresholds P _min = const ₁ and P _max = const ₂ , while in the case of crossing the adjusted ratio (16) thresholds P _min and P _max from top to bottom or from bottom to top, it is fixed the fact of detection of an object of ground equipment.

The value of the thresholds P _min and P _max was found from the analysis of relation (16) carried out for seismophone records on different days and under different weather conditions. It turned out that these limits, with sufficient accuracy for practice, are determined by the values const _n = 0.75 and const _b = 1.75. Thus, when an object of ground equipment approaches the seismic receiver from afar, relation (15) gradually turns into relation (16).

In FIG. 3a shows an example of detecting a URAL truck moving at a speed of 30 km/h at a distance from the geophone to the traverse point of 100 m; in fig. 3b - "BMP-3", moving at the same speed with the same traverse distance. Here, the bold dash-dotted lines show the levels Р _min and Р _max , and the bold stroke shows the equilibrium threshold.

If the readings (16) cross the thresholds P _max and P _min from top to bottom, then this means that the technical object starts a sufficiently powerful motor in the seismic receiver sensitivity zone and then starts moving. In FIG. 4 illustrates the situation when the T-72 tank started the engine 100 meters away from the traverse point and started moving at an average speed of 20 km/h.

If, during the passage of the object along the track, the readings (16) are located below the threshold Р _min , or after crossing any of the thresholds Р _min or Р _max remain between them, then the variant of skipping the useful signal will be implemented (non-detection of the object). This case is illustrated by the graphic (16) in FIG. 5, where the URAL truck was moving at a speed of 40 km/h at a distance from the seismic receiver to the traverse point of 250 m.

If, as a result of the analysis of the incoming seismic signal, all readings (16) are located above the threshold Р _max , then this means a corresponding imbalance of the seismophone energies in the low-frequency and high-frequency bands, which can be caused by some external factor. Since no level crossing has occurred, in this case the detection is ignored. Thus, the probability of false detection of an object is reduced.

In FIG. 6 and FIG. 7 shows the smoothed and amplified current values of the seismic signal energies in low-frequency bands, excited by moving ground objects of various classes in the ground. In FIG. 6a the graphic image refers to the tracked version of the BMP-3, moving at a speed of about 30 km / h, and in Fig. 6b - to the T-72 tank, moving at a speed of about 40 km / h. The traverse distance for these objects is 200 m. 7a refers to a UAZ-469 vehicle moving at a speed of about 25 km/h, and in FIG. 7b - to a URAL car moving at a speed of about 30 km/h. For these objects, the traverse distance is 100 m.

, тем больше будет это различие. Следовательно, при соблюдении баланса энергий сейсмофона в низкочастотной и высокочастотной полосах отношение текущих значений энергии смеси полезного сейсмосигнала и сейсмофона в низкочастотной полосе

к средней энергии сейсмофона

, вычисленной в последнем промежутке времени наблюдения в отсутствии подвижных объектов, определяемое соотношением Analysis of FIG. 6 and FIG. 7 shows that even despite the difference in the motion parameters of wheeled and caterpillar objects, it can be argued that the total energy of the mixture of useful seismic signal and seismophone concentrated in the low-frequency band for a caterpillar object is greater than the same energy for a wheeled object. Moreover, the smaller the ratio of the masses of objects of the "wheel" class and the "caterpillar" class

, the greater this difference will be. Therefore, if the balance of the seismophone energies in the low-frequency and high-frequency bands is observed, the ratio of the current values of the energy of the mixture of the useful seismic signal and the seismophone in the low-frequency band

to the average seismophone energy

, calculated in the last observation time interval in the absence of moving objects, determined by the ratio

, (17)

, (17)

for caterpillar objects will be greater than for wheeled objects with the same traverse distance. This determines the further operations of the proposed method.

3. Construct the following dimensionless ratio

, (18)

, (eighteen)

which can be called a criterion for distinguishing classes of objects, since the values of the current values (18), taking into account what was said above about the value (17), should tend to decrease before the traverse point and increase after this point. In this case, the sequence of minimum values (18), grouped in a small neighborhood around the traverse point for tracked objects, should be located below similar sequences of values (18) for wheeled vehicles.

The conducted studies of sufficiently representative statistics (100 records) of seismic signals from objects of the studied classes fully confirmed the conclusion formulated in the third paragraph. As an illustration, in FIG. 8 and 9 show graphic dependences (18) related to objects, smoothed and enhanced current values of seismic signal energies in low-frequency bands of which are shown in fig. 6 and FIG. 7. In FIG. 8 and FIG. 9, the thick horizontal line shows the statistically valid value of the classification threshold.

From the analysis of Fig. 8 and FIG. It follows from Table 9 that, in the average statistical sense, the sequence of minimum values (18) grouped in a small neighborhood around the traverse point for caterpillar objects is indeed below the classification threshold, while similar sequences of values (18) for wheeled vehicles are located above the threshold value.

с доверительной вероятностью

накрывается 10-ти процентным доверительным интервалом с нижней и верхней границами соответственно:

и

, то есть

.The analysis of 100 records of seismic signals from objects of ground equipment of the classes "wheel" and "caterpillar" (50 pieces "wheel", 50 pieces "caterpillar"), obtained with various parameters of the movement of objects, taking into account the features of their detection, as well as installation options seismic sensors on the ground, showed that the value of the statistical probability of correct classification of objects according to the enlarged classes "wheel" - "caterpillar"

with confidence

is covered by a 10 percent confidence interval with lower and upper bounds, respectively:

and

, that is

.

In FIG. 10 shows an enlarged block diagram of a device that implements the proposed classification method, where the following designations are introduced:

1 – three-component seismic receiver, three sensitivity axes of which are mutually orthogonal (SP 3D);

2, 4, 6 - the first, ..., third signal amplifiers (US);

3, 5, 7 - the first, ..., third band-pass filters (PF);

8 – analog-to-digital converter (ADC);

9 - microcontroller (MK);

10 - controlled switch for two positions (UP);

11, 12 - the first and second threshold devices (PU ₁ and PU ₂ );

13, 14 - the first and second decision devices (RU ₁ and RU ₂ ).

The device contains a three-component seismic receiver 1, three outputs of which are connected to the inputs of the first US 2, the second US 4 and the third US 6 signal amplifiers, respectively. The outputs of the first US 2 of the second US 4 and the third US 6 signal amplifiers through the corresponding first PF 3, second PF 5 and third PF 7 connected to the first, second and third inputs of the ADC 8, respectively. Three outputs of the ADC 8 are connected to the corresponding inputs of the microcontroller 9, the first and second outputs of which are connected to the first and second inputs of the controlled two-position switch 10, respectively. Moreover, the first output of the UE 10 is connected to the fourth input of the microcontroller 9 through the first threshold device 11 and the first decisive device 12 connected in series. classification devices. Three inputs of a three-component seismic receiver 1 are designed to supply three mutually orthogonal components of seismic signals.

The claimed device works as follows.

и выравнивающего коэффициента

. В случае принятия решения об обнаружении объекта с выхода первого решающего устройства 12 на четвёртый вход микроконтроллера 9 поступает информация в соответствующем формате. При этом в микроконтроллере 9 осуществляются операции (17) и (18). Одновременно с первого выхода микроконтроллера 9 на первый управляющий вход управляемого переключателя 10 поступает команда, по которой блокируется его первый выход и открывается второй его выход, с которого на вход второго порогового устройства 13 поступает последовательность отсчётов (18). Результат с выхода второго порогового устройства 13 поступает на вход второго решающего устройства 14, где в соответствии с поведением выходных отсчётов микроконтроллера 9 относительно величины порога классификации принимается решение о принадлежности обнаруженного объекта наземной техники к соответствующему классу. Одновременно с выхода второго решающего устройства 14 на пятый вход микроконтроллера 9 поступает информация в соответствующем формате, по которой с первого выхода микроконтроллера 9 на первый управляющий вход управляемого переключателя 10 поступит команда, по которой блокируется второй выход управляемого переключателя 10, открывается его второй выход и устройство классификации переходит в исходное состояние, то есть цикл обнаружения-классификации вновь будет осуществляться.In the initial state, the controlled switch 10 is in a position in which its second output is blocked, and the first one is open. The three inputs of the three-component seismic receiver 1 receive three mutually orthogonal components of the seismic signal, where they are converted into electrical signals. These electrical signals from the three outputs of the three-component seismic receiver 1 are fed to the corresponding inputs of the signal amplifiers 2, 4 and 6, where they are amplified to the required value. From the outputs of these amplifiers, the signals are fed to the corresponding inputs of band-pass filters 3, 5 and 7 each with a bandwidth equal to Δ f = 22 Hz and cutoff frequencies f _n = 10 Hz, f _in = 32 Hz. The filtered signals from the outputs of bandpass filters 3, 5 and 7 are fed to the three corresponding inputs of the analog-to-digital converter 8, where they are sampled and from its three outputs the discrete signals are fed to the corresponding inputs of the microcontroller 9. In the microcontroller 9, depending on the presence or absence of a ground-based object technique, the discretized components are processed in accordance with relations (9) - (16). The results in the form of current readings from the first output of the microcontroller 9 are fed to the input of the first threshold device 11, where their position relative to the thresholds P _max and P _min is controlled. The result of this control from the output of the first threshold device 11 is fed to the input of the first decision device 12, in which, in accordance with the behavior of the output readings of the microcontroller relative to the values of the thresholds P _max and P _min , a decision is made about the presence or absence of signs of a mobile object of ground equipment in the received seismic signal. If no signs of a moving object of equipment are found in the processed seismic signal, then the described detection cycle is repeated after a preliminary update of the data from the value of the average seismophone energy

and equalization factor

. If a decision is made to detect an object, information in the appropriate format is received from the output of the first decision device 12 to the fourth input of the microcontroller 9. In this case, operations (17) and (18) are carried out in the microcontroller 9. Simultaneously, from the first output of the microcontroller 9, a command is sent to the first control input of the controlled switch 10, by which its first output is blocked and its second output is opened, from which a sequence of readings (18) is fed to the input of the second threshold device 13. The result from the output of the second threshold device 13 is fed to the input of the second decision device 14, where, in accordance with the behavior of the output readings of the microcontroller 9 relative to the value of the classification threshold, a decision is made whether the detected object of ground equipment belongs to the corresponding class. Simultaneously, from the output of the second decision device 14, the fifth input of the microcontroller 9 receives information in the appropriate format, according to which a command will be sent from the first output of the microcontroller 9 to the first control input of the controlled switch 10, which blocks the second output of the controlled switch 10, opens its second output and the device classification goes to its original state, that is, the detection-classification cycle will again be carried out.

The implementation of this device does not cause difficulties, since all blocks, except for SP 3D 1 and UP 10, are quite often used in patents and engineering developments. A feature of the SP 3D 1 is that it is composed of three linear seismic sensors, the sensitivity axes are mutually orthogonal. Linear seismic sensors are well known, described in the technical literature, for example, in [5], or are used in appropriate devices, for example, in [6]. Options for implementing controlled switches for two positions 10 are proposed, in particular, in [7] and [8].

Sources of information

1. Patent 2202811 (RF). Device for detecting and classifying seismic signals. IPCG01V 1/16. Kryukov I.N., Ivanov V.A., Dyugovanets A.P., Afanasenko A.V. Application No. 2002113466/28 dated May 23, 2002. Published 04/20/2003 G.

2. Patent 2311665 (RF). Seismic device for detecting and classifying objects. IPCG01V 1/16. Sizov A.S., Strebkov D.A., Chelyshov S.Yu. Application No. 2006112075/28 dated 04/11/2006. Published 11/27/2007 G.

3. Patent 2236027 (RF). Seismic signal classification device. IPC G01V 1/16 . Kryukov I.N., Ivanov V.A., Matveev V.V. Application No. 2003118053/28 dated 06/19/2003. Published September 10, 2004

4. Patent 2697021 (RF). A method for ensuring the required probability of false alarms for a seismic signal classification device. IPCG01V 1/16,G08B 13/16; SPKG01V 1/001,G08B 13/1663,G01V 1/16. Anisimov V.I., Kravtsov A.V., Rusin P.V., Komyakov A.V. Application: No. 2018141621 dated 11/27/2018. Published 08.08.2019 G.

5. Yanchich V.V. Piezoelectric accelerometers based on a monolithic block with bending deformation / V.V. Yanchich // Foreign radio electronics. - 1996. - No. 9, pp. 63-64.

6. Patent 2204850 (RF). Seismic receiver. IPC G01V 1/16 . Share V.K., Kruglov A.K. Application No. 2002112574/28 dated May 13, 2002. Published May 20, 2003

7. Gorshkov, B.I. Radioelectronic devices / V.I. Gorshkov. - M.: Radio and communication, 1984. - 400 p.

8. Pukhalsky, G.I. Design of discrete devices on integrated circuits. Directory. / Pukhalsky G.I., Novoseltseva T.Ya. - M.: Radio and communication, 1990. - 304 p.

Claims (1)

A method for classifying moving objects of ground equipment by a seismic signal, including registration of a seismic signal excited by a moving object of ground equipment, filtering it in a selected frequency band, amplification and sampling, characterized in that in the absence of an object, the selected operating frequency range is divided into two bands - low-frequency and high-frequency - according to the condition of equality in them of the spectral densities of the seismophone, a periodic assessment of the average value of the energy of the seismophone in the selected band during the observation of the seismophone, as well as the value of the equalization coefficient of the energies of the seismophone in these bands as a value inverse to the ratio of the low-frequency band seismophone energy averaged over the observation time interval to the energy of the seismophone of the high-frequency band averaged over the same time interval; in the presence of a mobile object of ground equipment for the duration of the current time window, the energy value of the mixture of the useful seismic signal and the seismic phone in the low-frequency band is divided by the energy value of the mixture of the useful seismic signal and the seismic phone in the high-frequency band, the obtained quotient is multiplied by the last estimate of the leveling coefficient value and the current values are compared result with two thresholds - upper and lower; if these current values cross the lower or upper thresholds from top to bottom or from bottom to top, the fact of detection of an object of ground equipment is recorded, in the case when, after crossing these current values of the lower or upper thresholds, they remain between these thresholds, a variant of missing a useful seismic signal will be implemented - non-detection object, when all current values are above the upper threshold, the result is ignored; when a moving object is detected, the ratio of the current values of the energy of the mixture of the useful seismic signal and the seismophone in the low-frequency band to the average energy of the seismophone calculated during the last observation period in the absence of moving objects is calculated, and the result of this ratio is divided by the result of the ratio of the energy of the mixture of the useful seismic signal and the seismophone in the low-frequency band band to the same energy in the high frequency band; if the sequence of the minimum values of the result obtained, grouped in a small neighborhood around the traverse point, is below the statistically justified value of the classification threshold, then the object is classified as a “caterpillar”, if below this threshold, then it is classified as a “wheel”.

Images