RU2308055C2 - Method for forming of image of sea vessel contour according to radar surveillances - Google Patents

Abstract

FIELD: control of motion of sea vessels for provision of safety of sea traffic.

SUBSTANCE: a matrix is formed that contains echo-signals from the target and from the surface sea waves, whose columns serve as radar observation rules corresponding to the angular positions of the radar antenna, a bipolar matrix of wavelet-spectra is obtained, the elements of the like polarity that don't contain wavelet spectra of the echo-signals from the sea vessel hull are excluded from the matrix of the wavelet-spectra, the value of the binomization threshold is determined, binomization of the matrix of the wavelet-spectra is accomplished, the vessel image is separated by processing of the binomized matrix of the wavelet-spectra by a morphological filter.

EFFECT: the image of the contour of the sea vessel is formed according to two-dimensional (angle-distance) radar surveillances at a complete suppression of the interference component from the disturbed sea surface.

3 dwg

Description

The invention relates to the field of traffic control of marine vessels (hereinafter referred to as vessels) and is intended to form an image of the contour of a marine vessel from radar observations in order to expand the information base of ship traffic control systems (VTS) in the interest of ensuring maritime traffic safety.

One of the tasks facing the VTS, in conditions of high traffic flows, is to ensure traffic safety in the areas of the congestion of vessels: in straits, on fairways and on the routes of approach of vessels to port water areas. The main conditions for safe movement are: prevention of excessive rapprochement of vessels to a certain critical value that defines the security zone around the vessel and referred to as the “ship domain”, as well as possession of current information about the intentions of each vessel.

Traditionally, the ship's domain is the zone around the ship of a given radius, the shape and dimensions of which do not take into account the geometry, actual dimensions and current course of the ship. In this case, the ship's heading is determined by estimating the velocity vector from radar observations over several revolutions of the antenna, which indicates a delay in information about the direction and nature of the ship's movement.

In conditions of high traffic intensity in the places where the vessels are clustered, information on the geometric characteristics (length and width of the vessel) and the orientation of the hulls at the current time plays a significant role in monitoring the actual rapprochement of vessels (the shortest distance between the nearest points of the hulls).

In the VTS, such information can be obtained in the process of ship control by forming an image of the hull profile of each ship in the horizontal plane (hereinafter referred to as the ship’s contour) using two-dimensional (angle-distance) radar observations obtained by radar observation during one revolution of the antenna, which ensures VTS information on the length, width and course (angular position of the longitudinal axis of the contour) of each vessel, respectively, on the actual rapprochement of vessels in real time.

In general, radar observations generated during one revolution of the antenna at the output of the moving targets selection system (SAC) of a radar station (including an amplitude-to-digital converter - ADC) are an echo signal matrix whose columns are formed from radar observation lines. The radar line of observations for each angular position of the antenna is a sequence of discrete samples of the echo signal at the output of the ADC. A matrix of echo signals is generated for each marine vessel (hereinafter - the vessel), which falls into the VTS location zone.

The matrix of echo signals contains two main components: echo signals from the hull of the vessel and from the surface waves of the sea, which determine the information base for forming the image of the contour of the vessel.

Echo signals from the surface sea waves, which are interferences, exclude the possibility of directly obtaining information about the length, width and course of the vessel from the matrix of echo signals. For this, special methods are needed for imaging the contour of the vessel from radar observations.

Let us discuss the previously used definitions of the connected domain, binomization, and morphological image processing.

A connected area is a compact set of pixels, each pixel of which has at least one neighbor belonging to this set. A pixel is an element of an image, in this case, an element of a matrix.

Binomization is the transformation of a matrix with elements of arbitrary values into a matrix of binary values (elements of the matrix contain 1 or 0) using a given threshold.

Morphological processing of a binomized matrix using a two-dimensional morphological filter - the exclusion of the connected areas of the binomized matrix, the sizes of which are smaller than the dimensions of the two-dimensional matrix of the morphological filter.

There is a method of highlighting the boundaries of objects in grayscale images [1] using Sobel, Previt, Roberts and Canny filters, with which they form an image of the target contour from two-dimensional (angle-distance) radar observations by convolution of the observation matrix with a mask (two-dimensional matrix) of these filters .

The two-dimensional filter matrix has a central element. The convolution (filtering) procedure is carried out by sequentially moving the filter matrix so that its central element is alternately aligned with all elements of the echo matrix. In each of these positions, a summation of the elementwise multiplication of the elements of the filter matrix with the elements of the echo matrix that match the elements of the filter matrix is performed. If the summation result exceeds the threshold value set in advance, then the element of the resulting matrix (the image matrix of the target contour obtained by convolution of the matrix of echo signals with the filter matrix), whose coordinates (row number, column number) coincide with the coordinates of the central mask element, is assigned the value 1 otherwise 0.

The main disadvantage of this method is that the image of the contour of a marine vessel is distorted by interference from the agitated surface of the sea. This fact is explained by the fact that the radar echo signals, in addition to echo signals from the ship’s hull, also contain echo signals from the rough sea surface, which in this case are considered as interference. The use of Sobel, Previt, Roberts, and Canny filters reduces the level of interference by summing the elementwise multiplication of the filter matrix elements with the elements of the echo matrix, which coincide with the elements of the filter matrix, but do not completely exclude interference from the excited surface of the sea in the contour image of a marine vessel. Thus, the considered method does not contain procedures that exclude the listed disadvantages.

Thus, the indicated method for distinguishing the boundaries of objects in grayscale images cannot be used to form an image of the contour of a marine vessel with the complete suppression of the interference component from an excited sea surface.

A known method [2] is the selection of the contour in the radar image based on the wavelet transform, using which from a matrix containing echo signals from an agitated sea surface and echo signals from a radar target, the columns of which are radar observation lines corresponding to the angular positions of the radar antenna , get the image of the contour of the radar target by forming a bipolar matrix of the wavelet spectra of the columns of the matrix of echo signals.

The method under consideration uses the wavelet property, according to which they have both temporal and frequency localization, i.e. using wavelets, one can distinguish local spatial inhomogeneities of the signal (target contour), using the frequency properties of the wavelet as a filter with adjustable parameters. This method implements a discrete form (since the observations after the ADC are discrete) of a continuous wavelet transform in the form of a high-pass filter, with which a scalable differentiation procedure (selection of the target contour) is performed, which is implemented by convolution of columns (rulers) of the echo signal matrix with psi -wavelet function for a given scale (j) of the wavelet.

The psi function of a wavelet is a vector whose sum of elements is zero, since the elements of the first half of the vector have positive values, and the elements of the second are negative. During the convolution with elements of the radar observation line, the psi-function vector moves sequentially along the line so that the initial element of the vector alternately coincides with all elements of the line. In each of these positions of the psi-function vector, its elements are multiplied with the elements of the ruler, and the result of the multiplication is summed. The summation result is assigned to the element of the resulting ruler (the ruler obtained as a result of this wavelet transform), the coordinates of which coincide with the coordinates of the first element of the psi-function vector.

Thus, during the sequential movement of the psi-function along the ruler, for each position of the psi-function, the difference between two adjacent samples of the ruler (coinciding with the first and second halves of the elements of the psi-function vector) is established, the sizes of which increase with increasing j.

The result of this convolution of the psi function of the wavelet with echo signals from the radar target body is wavelet spectra in the form of unipolar pulses, the position of which on each line corresponds to the position of the echo signals from the reflecting element of the radar target body, then the set of resulting rulers forms a matrix of wavelet spectra , on which the said unipolar pulses form the outer contour of the casing of the radar target.

The result of the convolution of the psi-function of the wavelet with echo signals from the surface waves of the sea is the wavelet interference spectrum in the form of random bipolar pulses.

Thus, in general, the matrix of wavelet spectra of the echo signals, representing the sum of the wavelet spectra of the echo signals from the radar target body and surface waves, is a bipolar matrix of wavelet spectra.

With increasing parameter j, the sizes of the previously mentioned adjacent samples increase, i.e. their averaging effect on the noise component increases, the dispersion of the wavelet spectrum of which, as a result, decreases with increasing j. At the same time, the amplitudes of the wavelet spectra of the echo signals from the body of the radar target increase with increasing parameter j. Choosing the optimal value of the scale j, they achieve the best selection of the contour of the hull of the radar target. To reduce the level of interference to the wavelet spectra, an adaptive threshold is applied.

The main disadvantage of this method is that the image of the contour of a marine vessel is distorted by interference from the agitated surface of the sea.

This drawback is due to the fact that during the convolution of the psi-function of the wavelet with the columns (rulers) of the echo matrix, the averaging effect of the wavelet on the random component of the interference is expressed only in a decrease in the dispersion of the interference component and does not completely exclude it. The adaptive level of the threshold limitation of the elements of the wavelet spectra is focused on maintaining only a given level of false alarms and does not completely suppress the interfering component.

Moreover, according to the prototype method under consideration, elements of the same polarity that do not contain wavelet spectra of echo signals from the ship’s hull are not excluded from the bipolar matrix of the wavelet spectra of the echo signals, the binomization threshold at which the dimensions of the connected sets of binomized elements of the wavelet spectra is not set echoes from surface waves are smaller than the dimensions of a two-dimensional matrix of a morphological filter; binomization of the matrix of wavelet spectra of echo signals is not performed using the set threshold binomization, which excludes the elements of the wavelet spectra of the echo signals from the excited surface of the sea, using the indicated morphological filter by morphological processing of the generated binomized matrix of wavelet spectra, the wavelet spectra of the echo signals in the form of unipolar pulses whose position on each radar observation line are not obtained corresponds to the position of the echo signals from the reflecting element of the hull and which form the outer contour of the hull of the sea vessel.

Thus, the indicated method for isolating a contour in a radar image based on a wavelet transform cannot be used to form an image of the contour of a marine vessel while completely suppressing the interference component from the excited surface of the sea.

Consequently, the problem of forming an image of the contour of a marine vessel from two-dimensional (angle-distance) radar observations while completely suppressing the interference component from the excited sea surface remains unresolved.

The known method for isolating a contour in a radar image based on a wavelet transform, in terms of its technical nature, functionality and technical result achieved, is closest to the claimed invention for a method of forming an image of a contour of a marine vessel from radar observations and is further considered as a prototype method.

The basis of the invention is the creation of a method of forming an image of the contour of a marine vessel from radar observations with the complete suppression of the interference component from the excited surface of the sea. At the same time, it is taken into account that the main information elements provided by the vessel contour are the length, width and course (spatial orientation of the longitudinal axis of symmetry of the vessel contour) of the vessel.

The problem is solved in that in the method of forming an image of the contour of a marine vessel according to radar observations, in which a matrix is formed containing echo signals from the hull of the marine vessel and from surface sea waves, the columns of which are radar observation lines for each angular position of the antenna, which are a sequence of discrete samples of the echo signals, for each i-th line, where i = 1,2, ..., q is the number of the line, a bipolar matrix of the wavelet spectra of the echo signals is obtained by of the burst wavelet transform of the columns of the echo matrix, additionally elements of the same polarity that do not contain the wavelet spectra of the echo signals from the ship’s hull are excluded from the bipolar matrix of the wavelet spectra of the echo signals; spectra of echo signals from surface waves are smaller than the dimensions of a two-dimensional matrix of a morphological filter, binomization of the matrix of wavelet spectra of echo signals is performed using the updated binomization threshold, which excludes the elements of the wavelet spectra of the echo signals from the excited sea surface, using the indicated morphological filter, by morphologically processing the generated binomized matrix of wavelet spectra, the wavelet spectra of the echo signals in the form of unipolar pulses, the position of which on each radar line, are obtained observation corresponds to the position of the echo signals from the reflecting element of the hull and which form the outer contour of the hull of the ship but.

In the claimed method of forming the contour of a marine vessel by radar observations, the common essential features for him and for his prototype method are:

- the formation of a matrix containing echo signals from the hull of the ship and from the surface waves of the sea, the columns of which are radar observation lines for each angular position of the antenna, representing a sequence of discrete samples of echo signals, for each i-th line, where i = 1 , 2, ..., q is the number of the ruler;

- obtaining a bipolar matrix of wavelet spectra of the echo signals by continuous wavelet transform of the columns of the matrix of echo signals.

A comparative analysis of the essential features of the claimed method and the prototype method shows that the first, in contrast to the prototype method, has the following significant distinguishing features:

- the exclusion from the bipolar matrix of the wavelet spectra of the echo signals of elements of the same polarity that do not contain the wavelet spectra of the echo signals from the hull;

- setting the value of the binomization threshold at which the sizes of the connected sets of binomized elements of the wavelet spectra of the echo signals from surface waves are smaller than the dimensions of the two-dimensional matrix of the morphological filter;

- performing binomization of the matrix of the wavelet spectra of the echo signals using the set binomization threshold, which ensures the exclusion of the elements of the wavelet spectra of the echo signals from the excited sea surface,

- obtaining using the indicated morphological filter (by morphologically processing the generated binomized matrix) the wavelet spectra of the echo signals in the form of unipolar pulses, the position of which on each radar observation line corresponds to the position of the echo signals from the reflecting element of the ship’s hull and which form the outer contour of the marine hull vessel.

The set of features that ensure the achievement of a technical result:

- the formation of a matrix containing echo signals from the hull of the ship and from the surface waves of the sea, the columns of which are radar observation lines for each angular position of the antenna, representing a sequence of discrete samples of echo signals, for each i-th line, where i = 1 , 2, ..., q is the number of the ruler;

- obtaining a bipolar matrix of the wavelet spectra of the echo signals by continuous wavelet transformation of the columns of the matrix of echo signals;

- the exclusion from the bipolar matrix of the wavelet spectra of the echo signals of elements of the same polarity that do not contain the wavelet spectra of the echo signals from the hull;

- setting the value of the binomization threshold at which the sizes of the connected sets of binomized elements of the wavelet spectra of the echo signals from surface waves are smaller than the dimensions of the two-dimensional matrix of the morphological filter;

- performing binomization of the matrix of the wavelet spectra of the echo signals using the set binomization threshold, which ensures the exclusion of the elements of the wavelet spectra of the echo signals from the excited sea surface,

- obtaining using the indicated morphological filter (by morphologically processing the generated binomized matrix) the wavelet spectra of the echo signals in the form of unipolar pulses, the position of which on each radar observation line corresponds to the position of the echo signals from the reflecting element of the ship’s hull and which form the outer contour of the marine hull vessel.

The technical result from the application of the claimed method of forming an image of the contour of a marine vessel from radar observations is to obtain information about the geometric dimensions and the angular position of the longitudinal axis of the contour of the vessel with the complete suppression of the interference component from the excited surface of the sea in the image of the contour of the marine vessel.

This set of known and distinctive essential features is sufficient and necessary to achieve the claimed technical result.

Based on the foregoing, we can conclude that the set of essential features of the claimed invention has a causal relationship with the achieved technical result, i.e. thanks to this combination of essential features of the invention, it has become possible to solve the problem.

Therefore, the claimed invention is new, has an inventive step, i.e. it does not explicitly follow from the prior art and is suitable for industrial use.

The essence of the claimed method of forming an image of the contour of a marine vessel from radar observations is illustrated by the drawings:

figure 1. The appearance of the matrix of echo signals;

- figure 2. The appearance of the binary matrix of the wavelet spectra of the echo signals from the hull of the vessel after processing with a morphological filter.

- figure 3. The contour of the ship.

The claimed method of forming an image of the contour of a marine vessel from radar observations is carried out by step 1 of forming a matrix containing echo signals from the hull of the marine vessel and from surface sea waves (Fig. 1), the columns of which are radar observation lines for each angular position of the antenna, which are a sequence discrete samples of echo signals for each i-th line, where i = 1,2, ..., q is the number of the line; step 2 of obtaining a bipolar matrix of the wavelet spectra of the echo signals by continuously wavelet transforming the columns of the matrix of echo signals; operation 3, exceptions from the bipolar matrix of the wavelet spectra of the echo signals of elements of the same polarity that do not contain the wavelet spectra of the echo signals from the hull; operation 4 of setting the binomization threshold value, in which the sizes of the connected sets of binomized elements of the wavelet spectra of the echo signals from surface waves are smaller than the dimensions of the two-dimensional matrix of the morphological filter; step 5 of performing binomization of the matrix of wavelet spectra of the echo signals using the set binomization threshold, which excludes elements of the wavelet spectra of the echo signals from the excited sea surface; step 6 of obtaining using the indicated morphological filter (by morphologically processing the generated binomized matrix) the wavelet spectra of the echo signals in the form of unipolar pulses (Fig. 2), the position of which on each radar observation line corresponds to the position of the echo signals from the reflecting element of the ship’s hull and which form the outer contour of the hull of the marine vessel (figure 3).

Implementation of the claimed method of imaging the contour of a marine vessel from radar observations is as follows.

In accordance with the rules for safe navigation of vessels, the minimum distance between vessels cannot be less than their ship domains, i.e. there is always a guaranteed distance between vessels that is not occupied by vessels. At the same time, the SDC “captures the target” in such a way that the mark from the target (vessel) is always in the center of the two-dimensional sample (angle-distance) of the radar observations. The sample formed using the ADC and corresponding to a certain angular position of the antenna, in this case is a fragment of the radar line of observations (hereinafter - the line).

Therefore, the matrix of echo signals (operation 1), obtained during one revolution of the radar antenna, contains in the central part elements whose values are equal to the sum of the echo signals from the ship's hull and surface sea waves, the rest of the matrix elements contain only echo signals from surface waves of the sea.

From the lines corresponding to each angular position of the antenna, an echo matrix is formed

- порядковый номер отсчета АЦП на линейке; n - максимальное число отсчетов АЦП на линейке; i=1,2,...,q - номер линейки, q - выбранное число линеек в матрице Z.where Z ₀ (k, i) is the matrix of echo signals from the hull of the ship; ε (k, i) is the matrix of echo signals (interference) from the excited surface of the sea;

- serial number of the ADC reading on the line; n is the maximum number of ADC readings on the line; i = 1,2, ..., q is the ruler number, q is the selected number of rulers in the matrix Z.

For each i-th line, a wavelet spectrum is formed using the discrete form of a continuous wavelet transform [3]

- координата положения пси-функции на линейке; h - число элементов вектора пси-функции ψ₀.where ψ ₀ is the basic psi-function of the wavelet; j is the scale parameter;

- the coordinate of the position of the psi function on the ruler; h is the number of elements of the vector of the psi function ψ ₀ .

Continuous wavelet transform is a scalable differentiation procedure implemented by convolution of a column of matrix Z (ruler) and psi function ψ ₀ , during which the difference between two adjacent samples is established, the sizes of which increase with increasing j, and the weighting factors of sample elements during convolution are determined by the psi-function ψ ₀ . Moreover, with increasing parameter j, the sizes of adjacent samples increase, i.e. their averaging effect on the noise component increases, the dispersion of the wavelet spectrum of which, as a result, decreases with increasing j. At the same time, the amplitudes of the wavelet spectra of the echo signals from the hull of the vessel increase with increasing parameter j. In this case, the envelope of the wavelet spectrum of the echo signals from the ship’s hull is characterized by a change from zero to saturation (maximum) and again to zero — the larger the parameter j, the wider the saturation region. The saturation region (for given ψ ₀ and j) is a measure of the uncertainty of the coordinate on the radar line of the wavelet spectrum of the echo signal from the ship's hull.

The result of the convolution (2) of the psi function ψ ₀ with the columns of the matrix Z ₀ (k, i) is the wavelet spectra of the echo signals from the ship's hull in the form of negative pulses with an amplitude U _c and a width equal to the saturation region, the position of which on each i The ith line corresponds (taking into account the measure of uncertainty) to the position of the leading edge of the rectangular pulse of the echo signal from the hull of the vessel irradiated by the radar. In the future, a lot of pulses U _c determines the geometry of the contour of the vessel.

The result of the convolution (2) of the psi function ψ ₀ with the columns of the matrix ε (k, i) of echo signals from the excited sea surface is the wavelet interference spectrum in the form of bipolar elements of the matrix C _i (b, j), the value of which increases with increasing amplitude echo signals from the excited surface of the sea and decreases, as mentioned above, with increasing scale parameter j.

(операция 2).Further, the vectors of the wavelet spectrum (2) for the selected j are grouped into a bipolar matrix C _j (b, i); i = 1,2, ..., q;

(operation 2).

Due to the fact that the positive values of the elements of the matrix With _j (b, i) do not carry information about the contour of the vessel (pulses U _c have a negative polarity) and are interfering components of the wavelet spectrum, then the matrix With _j (b, i ) ^- , in which the positive elements (interfering components) of the matrix C _j (b, i) are replaced by zero values (operation 3).

This technique is one of the distinctive elements of the method of forming an image of the contour of a marine vessel based on reducing the connectivity of the elements of the matrix of wavelet spectra. Its essence lies in the fact that after the exclusion of positive elements from the matrix C _j (b, _i ), the connectivity of the bipolar elements of the interference wavelet spectrum is destroyed, since the positive elements of the interference component are replaced by zero values - the zero elements destroy the connectivity of the interference wavelet spectra.

At the same time, on all the lines, negative pulses U _c , which carry information about the geometry of the hull of the vessel, are preserved and their connectivity is not broken.

A specific value of the scale j for a given psi-function ψ ₀ is selected from the condition of reaching the maximum ratio of the amplitudes of the pulses U _c to the dispersion of the wavelet interference spectrum.

, где abs(^.) - оператор абсолютного значения.Further, for brevity, the matrix of wavelet spectra is denoted by

where abs ( ^. ) is the absolute value operator.

, которая содержит только вейвлет-спектры помех. Так как наблюдения (1) сформированы таким образом, что эхо-сигналы от корпуса судна располагаются в центральной части матрицы Z (k,i), то, соответственно, имеется возможность выбрать ту часть матрицы вейвлет-спектров, которая заведомо не содержит вейвлет-спектры эхо-сигналов от корпуса судна.From the matrix of wavelet spectra C, that part of it is selected

which contains only wavelet interference spectra. Since observations (1) are formed in such a way that the echo signals from the ship’s hull are located in the central part of the matrix Z (k, i), it is therefore possible to choose the part of the matrix of wavelet spectra that obviously does not contain wavelet spectra echo signals from the hull.

Note that the statistical characteristics of the surface sea waves are determined by the intensity of the wind and the characteristics of the water area, which do not change within the local area that forms the radar observations recorded in the matrix of echo signals. Therefore, the selected matrix C has a statistical representativeness of the interference wavelet spectrum for the entire matrix C of the wavelet spectra of the echo signals.

репрезентативны по отношению ко всей матрице вейвлет-спектров С, то установление порога р на матрице

справедливо также и по отношению к матрице вейвлет-спектров помех С.Another distinctive element of the method of forming an image of the contour of a marine vessel based on reducing the connectivity of the elements of the interference component of matrix C is the determination of the binomization threshold p (step 4), which is adaptive to the level of the interference wavelet spectrum, to the type of psi function ψ ₀ , the magnitude of the scale j, and the dimensions of the two-dimensional morphological filter. Since the statistical characteristics of the wavelet spectrum of the matrix interference

are representative with respect to the entire matrix of wavelet spectra C, then the establishment of the threshold p on the matrix

also true with respect to the matrix of wavelet spectra of interference C.

, оставшихся после операции 3, получает нулевые значения в процессе биномизации матрицы

на уровне р. Указанная процедура обозначается оператором ВThe decrease in connectivity based on the application of the threshold p is due to the fact that some of the nonzero interference components of the matrix

remaining after operation 3 gets zero values in the process of binomizing the matrix

at the level of p. The indicated procedure is indicated by operator B

,

,

- элемент матрицы

.Where

,

,

- matrix element

.

с помощью двумерного морфологического фильтра (морфологический фильтр [4]).The procedure for setting the binomization threshold value, in which the sizes of the connected sets of binomized elements of the wavelet spectra of the echo signals from surface waves are smaller than the dimensions of the two-dimensional matrix of the morphological filter (step 4), includes morphological processing of the binary matrix of the interference wavelet spectrum

using a two-dimensional morphological filter (morphological filter [4]).

The indicated procedure is denoted by the operator M

при выбранных размерах c×c двумерной матрицы морфологического фильтра зависят от величины порога р. Чем больше порог р, тем меньше связность множеств биномизированных элементов вейвлет-спектров помех и, соответственно, меньше значения элементов матрицы

, и, наоборот, чем меньше порог р, тем больше связность множеств биномизированных элементов вейвлет-спектров помех и, соответственно, больше значения элементов матрицы

.Since the amplitude of the wavelet interference spectrum is random (due to the random nature of the echo signals from the excited surface), the values of the matrix elements

at the selected sizes c × c, the two-dimensional matrix of the morphological filter depends on the threshold p. The larger the threshold p, the less connected the sets of binomized elements of the wavelet interference spectra and, accordingly, the smaller the values of the matrix elements

, and, conversely, the smaller the threshold p, the greater the connectivity of the sets of binomized elements of the wavelet interference spectra and, accordingly, the greater the value of the matrix elements

.

The optimal value of the threshold p is determined in the course of solving the optimization problem

- целевая функция оптимизационной задачи;

- элементы матрицы

. При оптимальном значении порога, равном р^*, размеры связных множеств биномизированных элементов вейвлет-спектров эхо-сигналов от поверхностного волнения меньше размеров двумерной матрицы морфологического фильтра, что обеспечивает полное исключение элементов вейвлет-спектра помех из матрицы

.Where

- objective function of the optimization problem;

- matrix elements

. At the optimal threshold value equal to p ^* , the sizes of the connected sets of binomized elements of the wavelet spectra of the echo signals from surface waves are smaller than the dimensions of the two-dimensional matrix of the morphological filter, which ensures the complete exclusion of elements of the wavelet interference spectrum from the matrix

.

Then, in step 5, similarly to rule (3), the matrix of wavelet spectra C is binomized using the obtained binomization threshold p ^* .

и С размеры связных множеств элементов вейвлет-спектров помех биномизированной матрицы С также будут меньше размеров двумерного морфологического фильтра, что обеспечит полное подавление помехи матрицы С в результате морфологической обработки элементов ее вейвлет-спектра. При этом множества элементов матрицы С^В(р^*)=В{С(р^*)}, содержащие компоненту U_с вейвлет-преобразования эхо-сигналов от корпуса судна, в отличие от вейвлет-спектра помех, сохраняют связность. Следовательно, в ходе морфологической обработка (операция 6), выполненной морфологическим фильтром, формируется матрицаMoreover, due to the previously substantiated representativeness of the statistical characteristics of the wavelet spectra of matrix noise

and C, the sizes of the connected sets of elements of the wavelet interference spectra of the binomized matrix C will also be smaller than the dimensions of the two-dimensional morphological filter, which will ensure complete suppression of the interference of the matrix C as a result of the morphological processing of the elements of its wavelet spectrum. Moreover, the sets of matrix elements C ^B (p ^* ) = B {C (p ^* )} containing the component U _{with the} wavelet transform of the echo signals from the ship's hull, unlike the wavelet interference spectrum, retain connectivity. Therefore, during the morphological processing (operation 6) performed by the morphological filter, a matrix is formed

elements of which form an image of the contour of a marine vessel.

Thus, the image formation of the contour of a marine vessel is performed with the complete suppression of the echo signals of the interfering component from the surface sea waves.

Currently, a device that implements the claimed method is in the process of prototyping. The prototyping was performed in relation to the coastal radar station of the circular review of the VTS Center of the port of Vladivostok.

Radar and ship specifications:

- the width of the antenna pattern - 0.39 °;

- sampling frequency of the ADC of the echo signal at the output of the radar video detector - 75 MHz (linear sampling interval of observations - 2 m);

- dimensions of the vessel: length - 160 m, width - 28 m;

- vessel heading - 45 °;

- distance to the ship - 1000 m;

- the angular pitch of the antenna between two adjacent radar arrays is 0.03 °.

We form the layout of the matrix of echo signals (1) in the form of the sum of the main components: echo signals from the ship's hull (Z ₀ ) and reflection from surface sea waves (ε).

The probability density distribution of the effective scattering area (EPR) of these reflectors is described [5]:

- exponential law for EPR (S ₀ ) echo signals from the hull

- среднее значение ЭПР корпуса судна в угловом секторе диаграммы направленности антенны;Where

- the average value of the EPR of the hull in the angular sector of the antenna pattern;

- the Weibull distribution for the EPR (S _ε .) of surface sea waves

where c '= 1 / b' is the shape parameter, G is the median of the distribution, determined by the intensity of the waves.

=1,0, c'=0,57, G=0,8 и, принимая во внимание

, сформируем [5] в указанных соотношениях матрицу эхо-сигналов от корпуса судна и поверхностного волнения моря (фиг.1). Как следует из фиг.1 эхо-сигналы от корпуса судна существенно искажены высоким уровнем помех от взволнованной поверхности моря.Choose model parameters

= 1.0, c '= 0.57, G = 0.8 and, taking into account

, form [5] in the indicated ratios the matrix of echo signals from the hull and surface sea waves (Fig. 1). As follows from figure 1, the echo signals from the hull are significantly distorted by a high level of interference from the agitated surface of the sea.

We form the matrix of wavelet spectra C at a selected scale j = 3, which ensures the maximum ratio of the pulse amplitude U ^c to the dispersion of the interference wavelet spectrum and the psi function ψ _{0 of the} Daubechies wavelet 'db1' [3], which has the maximum linear resolution, which allows minimizing the saturation region mentioned above, which is a measure of the uncertainty of the coordinates of the position of the contour of the vessel.

We choose the morphological filter bwmorph, type = 'majority' [4] with the dimensions of the two-dimensional filter c × c = 3 × 3 and the parameter f = 5, which has the maximum resolution, which also ensures high accuracy in the formation of the image of the contour of the vessel.

, сформированной из начальных фрагментов линеек матрицы С, которые заведомо содержат только вейвлет-спектры помех, установлено для данной макетной ситуации оптимальное значение порога биномизации р^*=0,335.As a result of modeling solutions to the optimization problem (5) on the matrix

formed from the initial fragments of the lines of the matrix C, which obviously contain only the wavelet spectra of interference, the optimal value of the binomization threshold p ^* = 0.335 was established for this mockup situation.

The application of the set threshold p ^{* of} binomization of the wavelet spectra matrix to the entire matrix C and subsequent morphological processing by the selected filter allows to completely suppress the noise background and at the same time generate the binary matrix C ^BM (figure 2). Matrix C ^BM is a figure formed of single elements, which in the horizontal plane form the contour of a marine vessel (figure 3).

Since the matrix C ^BM represents a figure formed of single elements, the specified figure in the horizontal plane has both external and internal boundaries, the distance between which is equal to the width of the saturation region mentioned above (pulse width U _c ) and the corresponding measure of uncertainty, which for given ψ ₀ and j for a given mock situation does not exceed 4 m.

The part of the ship’s hull closest to the radar is a continuous reflecting surface of the side of the ship, the contour image of which in the horizontal plane is represented by continuous curves 1 and 2. The far part of the ship’s hull is in the radar shadow zone, therefore, when prototyping it is provided that the sources of echo formation in this In the case, there are separate reflectors irradiated with a radar, imitating a bulwark and rail rails located on the upper edge of the far side. This explains the fragmentation of the longboard contour of the vessel (curves 3 and 4).

The absence of fragments of the contour of the far side in the stern of the vessel imitates the screen action of the stern superstructures, the contour of which 5 merges with the image of contour 2 in the stern of the vessel.

It should be noted that the presence of internal and external borders of the figure formed by the individual elements of the matrix C does not distort the shape of the contour of the vessel and does not introduce errors in determining the size and course of the vessel. So, curves 1 and 3 or 2 and 4, taken in pairs, form an image of the contour of the vessel, allowing you to determine the size and course of the vessel. Moreover, the specified uncertainty affects only the determination of the exact distance to the vessel, the error of determination of which in this model situation does not exceed 2 m (half of the uncertainty measure), which is 0.2% (the distance to the vessel in this case is 1000 m).

The image of the ship's contour has a stepped view, due to the discrete nature of the formation of observation lines using the ADC.

A further decrease in the angular pitch of the radar antenna and an increase in the sampling frequency of the ADC allows to increase the number of elements in the echo-signal matrix and thereby reduce the size of the steps and bring the generated image of the ship's contour closer to the true profile of its hull in the horizontal plane.

Prototyping confirms the possibility of forming an image of the contour of the vessel and determining on its basis the dimensions (length - 160 m, width - 28 m) and the course K of the vessel, which coincides with the longitudinal axis of symmetry of contour 6 and in this case is 45 ° relative to the direction to the North. In this case, the background noise from the excited surface of the sea in the image of the contour of the vessel is completely suppressed.

Thus, the obtained results confirm the achievement of the technical result from the application of the claimed method for imaging a contour of a marine vessel from radar observations, which consists in obtaining information about the geometric dimensions and angular position of the longitudinal axis of the contour of the vessel with the complete suppression of the interference component from the excited surface of the sea on the contour image of the marine vessel .

Information sources

1. Rudakov P.I., Safonov I.V. Signal and image processing. MATLAB 5.x / Under the general. Ed. V.G. Potemkina. M .: DIALOGUE-MIFI. 2000.416 s. - ANALOGUE.

2. Yang Li, Yuan Xin. An Edge Detector Based On Wavelet Transform For SAR Image.// Trans. Nanjing Univ. Aeron and Astron. 2001. Vol.8, N 1. P.75-80 - PROTOTYPE.

3. Dyakonov V.P. Wavelets. From theory to practice. M .: SOLON-R, 2002.448 s.

4. Dyakonov V.P. MATLAB. Signal and image processing. Special reference. SPb .: Peter. 2002.608 p.

5. Dorozhko VM, A simulation model of a radar echo signal. // Far Eastern Mathematical Journal, 2000, vol.2.1. S.98-113.

Claims (1)

The method of forming an image of the contour of a marine vessel from radar observations, which consists in the fact that they form a matrix containing echo signals from the hull of the marine vessel and from surface sea waves, the columns of which are radar observation lines for each angular position of the antenna, which are a sequence of discrete echo samples -signals, for each ith line, where i = 1,2, ..., q is the line number, a bipolar matrix of the wavelet spectra of the echo signals is obtained by continuous wavelet transform columns of the echo signal matrix, characterized in that elements of the same polarity that do not contain the wavelet spectra of the echo signals from the ship’s hull are excluded from the bipolar matrix of the wavelet spectra of the echo signals, the binomization threshold value is set at which the dimensions of the connected sets of binomized wavelet elements spectra of echoes from surface waves are smaller than the dimensions of a two-dimensional matrix of a morphological filter, binomization of the matrix of wavelet spectra of echoes is performed using the set threshold of anomization that ensures the elimination of the elements of the wavelet spectra of the echo signals from the excited surface of the sea, using the indicated morphological filter, by morphologically processing the generated binomized matrix of the wavelet spectra, the wavelet spectra of the echo signals in the form of unipolar pulses, the position of which on each observation radar line corresponds the position of the echo signals from the reflecting element of the hull and which form the outer contour of the hull of the marine vessel.

Images