RU2256169C1 - Method and device for investigating object in dissipated and/or passed radiation - Google Patents

Abstract

FIELD: object investigations in dissipated or passed radiation.

SUBSTANCE: broad irradiating beam covering entire object to be investigated is used for inspecting this object with passed or dissipated penetrating radiation. Novelty is use of beam with time-variable radiation intensity distribution in its cross-sectional area ensured by moving screen heterogeneous for used radiation across beam. Resultant definition depends on penetrating radiation measurement quantization step upon interacting with object at essential increase in effectiveness of source energy use.

EFFECT: enhanced scan speed due to effective use of source power or reduced power of source at same scan speed.

19 cl, 19 dwg

Description

The present invention relates to means for non-destructive testing, introscopy or inspection of humans and vehicles for the detection of foreign objects and is intended to obtain an image of a given area of the object using transmitted or scattered penetrating radiation.

A known method and device for inspection of a person in scattered X-ray radiation [1] for the detection of foreign objects on the human body (drugs, explosives, etc.). An x-ray pointed beam scans the surface of the human body and causes scattered radiation, which is detected by a large-sized scattered radiation detector. This registered scattered radiation is displayed and visualized. The contrast in the image is determined by the dependence in the scattering of x-ray radiation on substances with different atomic numbers. The lower the atomic number of a substance, the better it scatters the x-ray radiation. Thus, it is possible to detect substances on the human body that do not give good contrast when they are scanned on a fluorographic-type apparatus. However, this approach does not allow you to look deep enough into the object under study. The spatial resolution of the rendered image is determined by the width of the sharply directed scanning x-ray. An oversized x-ray beam is thus formed from a wide conical beam by collimation, due to which most of the initial wide beam is “cut off” by the collimator. The energy of the original x-ray source is used inefficiently.

A known method and device [2], which combines the advantages of research in scattered and transmitted radiation. In this device and method, the object under study is irradiated with a highly directed x-ray beam and both transmitted and scattered radiation are recorded by scattered and transmitted radiation detectors. Data obtained in both scattered and transmitted radiation are subjected to visualization and analysis. Thus, it is possible to detect well both objects with a large atomic number (metal weapons) and objects with a small atomic number (drugs, explosives). The spatial resolution in the visualized attenuation and scattering images is determined by the width of the sharpened x-ray scanning beam. An oversized x-ray beam is thus formed from a wide conical beam by collimation, due to which most of the initial wide beam is “cut off” by the collimator. The energy of the original x-ray source is used inefficiently.

There is also a known method and device according to the application for the invention [3], in which both scattered and transmitted x-ray radiation are used to inspect the object. Moreover, to irradiate an object, either a source of penetrating radiation is used to irradiate an object in many of its parts, or a source of penetrating radiation is used to irradiate an object with many beams. These technical solutions use a modulation system to encode each part of the irradiated object with such a time model of intensity variation that, by decoding the signal, each signal component can be uniquely associated with its part of the object.

In the embodiment described in this application for invention, coding is carried out on the basis of irradiation of various parts of the subject with beams modulated in frequency. Each beam irradiating its part of the object has its own modulation frequency. Thus, by separating the observed signal after interacting with the object with frequency filters, it is possible to separate the signals coming from different parts of the object.

The final resolution corresponds to the use of a single narrow beam.

This approach allows us to quite simply solve the problem of frequency coding of a flat fan beam, for example, due to the rotation of a disk with concentric holes located in the beam section, which is proposed in this application for invention [3], but makes it difficult to solve the problem of coding a conical beam.

In the aforementioned application for invention [3], the use of a fan beam also implies the movement of the studied object in the process of its study, since it is impossible to completely illuminate the entire object with a fan beam.

The method and device according to the application [3] are closest to the proposed ones.

The present invention is aimed at obtaining a technical result, which consists in increasing the scanning speed due to more efficient use of the power of the source of penetrating radiation, or in reducing the power of the source while maintaining the scanning speed. In addition, they simplify the design of the device and reduce its cost due to the use of non-segmented detectors for both transmitted and scattered radiation.

The proposed method for obtaining an image of a given region of an object, as well as the above-mentioned known method [3], includes scanning in time a given region of an object by a beam of penetrating radiation and measuring in time the intensity of penetrating radiation after interacting with an object to obtain a signal corresponding to this intensity, as well as the subsequent processing of this signal.

To achieve these technical results in the proposed method, in contrast to the closest known to him, carry out simultaneous irradiation with a beam of prinimaya radiation of the entire specified area of the object. At the same time, a time scan of a given region of the object is carried out by changing the time distribution of the radiation intensity in the section of the irradiating beam so that the received signal is a convolution and / or Radon transform over the image of the object. Processing these measurement results involves performing reverse convolution and / or inverse Radon transforms on them. The result is a restored undistorted (restored) image of a given area of the object.

The specified convolution should preferably be a cyclic convolution, and the reverse convolution, respectively, a reverse cyclic convolution.

A specific type of said convolution can be selected according to a given criterion, for example, to maximize the signal-to-noise ratio in the resulting image.

In particular, said convolution may be a convolution with a function having flat side lobes and one maximum in its discrete autocorrelation function.

The measurement of the intensity of penetrating radiation after interacting with an object can be carried out discretely in time.

The penetrating radiation used may be, for example, x-ray radiation.

The proposed device for obtaining an image of a given region of an object, as well as the one closest to it, known from [3], contains a source of penetrating radiation and at least one detector for measuring the intensity of penetrating radiation after interacting with the object and generating a signal corresponding to this intensity.

To achieve the above technical results in the proposed device, in contrast to the closest known to it, the source of penetrating radiation is configured to form a beam for simultaneous irradiation of the entire specified area of the object. In addition, the device includes a scanner for changing over time the distribution of the radiation intensity in the cross section of the specified beam in such a way that the signal generated by at least one detector in time, corresponding to the intensity of the penetrating radiation after interacting with the object, is the result of convolution and / or Radon transform above the image of the object. The device also comprises a processor for performing the inverse mathematical transformation on the signal corresponding to the intensity of the penetrating radiation after interacting with an object formed by at least one detector, i.e. to perform the inverse convolution and / or inverse Radon transform operation.

The scanner may contain a screen inhomogeneous for the penetrating radiation and means for moving this screen in at least one direction.

The specified screen inhomogeneous for penetrating radiation may be a drum having a one-dimensional pattern of length M · N, repeated along the height of the drum N times, with a shift by a length M for each subsequent repeat in height. Moreover, the means for moving the screen in at least one direction is configured to rotate the specified drum around its axis.

The means for moving the screen included in the scanner inhomogeneous for penetrating radiation can be configured to move this screen in two directions.

In this case, the screen inhomogeneous for the penetrating radiation can be a drum with a two-dimensional pattern deposited on its surface, cyclically repeating along the height and circumference of the drum, and the means for moving the screen in two directions is configured to rotate this drum around its axis and move along it .

In the same case, the screen inhomogeneous for penetrating radiation can be a sphere having a two-dimensional pattern on its surface, applied cyclically relative to the orthogonal axes passing through its center, and the means for moving the said screen in two directions is made with the possibility of rotation of this sphere around the indicated axes.

In the same case, the screen inhomogeneous for penetrating radiation can be a drum with a one-dimensional pattern printed cyclically around its circumference, and the means for moving the screen in two directions is configured to rotate this drum simultaneously around its axis and around an axis orthogonal to it.

In the same case, the screen inhomogeneous for penetrating radiation can be a disk, on the surface of which a one-dimensional pattern is applied relative to the center of the disk, and the means for moving the said screen in two directions is made with the possibility of rotation of this disk around its center and rotation of the center of the disk around the circle, lying in the plane of the disk (planetary motion).

The scanner may also contain two screens inhomogeneous for the penetrating radiation and a means for moving each of the screens in its direction.

In this case, one of the two screens inhomogeneous for the penetrating radiation can be a drum with a one-dimensional pattern cyclically applied around its circumference.

In the same case, screens that are heterogeneous for penetrating radiation can be two drums placed one in another with orthogonally oriented axes and one-dimensional patterns cyclically applied around their circles, and the means for moving these screens can rotate these drums around these axes.

In the same case, screens that are inhomogeneous for penetrating radiation can consist of two drums coaxially placed into each other drums with one-dimensional patterns cyclically applied around their circumference, the patterns of these drums being tilted relative to their common axis in opposite directions by an angle of 45 °, and the means for the movement of these screens is made with the possibility of rotation of these drums around their common axis.

The source of penetrating radiation may be a source of x-ray radiation.

The meaning of the well-known mathematical concepts used is explained below:

“Convolution” (direct aperiodic convolution).

In direct aperiodic convolution, instead of the value of the aperiodic function I (x) at point x, its weighted average value C (x) is used in the range from minus to plus infinity, while another function H (x) is used as the "weight". In analog form, such a convolution is described by the convolution integral, in a discrete form, such convolution is described by the sum.

“Reverse convolution” (inverse aperiodic convolution). In reverse aperiodic convolution, instead of the value of the aperiodic function C (x) at point x, its weighted average value I (x) is used in the range from minus to plus infinity, while another function H ^-1 (x) is used as the “weight”.

The function H ^-1 (x) is chosen in such a way as to compensate for the effect of direct aperiodic convolution, that is, to obtain an undistorted value of the function. In analog form, such a convolution is described by the convolution integral, in a discrete form, such convolution is described by the sum. Such a procedure is also called deconvolution.

“Loop convolution”

In cyclic convolution, instead of the value of the cyclic (periodic) function I (x) at point x, its weighted average value C (x) is used in the range of the repetition period, while another cyclic function with the same repetition period is used as the "weight".

“Reverse loop convolution”.

In the reverse cyclic convolution, instead of the value of the cyclic function C (x) at the point x, its weighted average value I (x) is used in the range of the repetition period, and another function H ^-1 (x) is used as the “weight”.

The function H ^-1 (x) is chosen in such a way as to compensate for the effect of direct cyclic convolution, that is, to obtain an undistorted value of the function. In analog form, such a convolution is described by the convolution integral, in a discrete form, such convolution is described by the sum.

“Direct Radon Transformation.”

The direct Radon transform reduces to the translation of the function I (x, y) from the Cartesian coordinate system to the coordinate system of the angular projections of this function P (Q, s). where Q is the angle at which the projection of the function is obtained, s is the coordinate along the projection.

In analog form, such a transformation is described by the integral equation, in discrete form by the sum.

A clear example of the direct Radon transform is the acquisition of x-ray images of an object from different directions around the object.

“The inverse Radon transform."

The inverse Radon transform reduces to the translation of the function P (Q, s) from the coordinate system of the angular projections into the Cartesian coordinate system I (x, y). In analog form, such a transformation is described by a complex integral equation, in discrete form - by a specific sum.

A good example of the inverse Radon transform is the acquisition of a tomographic image of a patient in medical computed tomography based on a set of X-ray projections obtained from various angles.

To further clarify the essence of the proposed inventions, we will consider how classical scanning is converted in these technical solutions.

Classical scanning is performed with a narrow beam, when the beam position fully characterizes the position of the part of the object being irradiated. In this case, it is not necessary to perform any coding of information coming from the part of the object being irradiated at the moment, since the position of the beam completely determines the part of the object being irradiated at the moment.

To increase the efficiency of using the power of a source of penetrating radiation, it is possible to expand the beam width. If you scan an object with an expanded beam, this will lead to the fact that in a unit time the object will be irradiated with a large number of quanta. However, the scan result will be changed. The nature of the changes will be characterized by the convolution of the desired image (scanned by a narrowly focused beam) and the distribution of the radiation intensity in the beam cross section. Nevertheless, if the distribution of the radiation intensity in the beam cross section is known, then it is possible to perform the reverse convolution and restore the undistorted desired image.

Further, this approach can be further generalized, based on scanning an object under study with an expanded beam, expanding the beam to sizes equal to the object being studied, or larger. In this case, the object per unit time will be irradiated with even more quanta, however, the nature of the changes in the desired image will not change and will also be characterized by the convolution of the desired image and the distribution of the radiation intensity in the cross section of the scanning beam. Consequently, the unchanged desired image can be restored by performing reverse convolution if the distribution of the radiation intensity in such a wide scanning beam is known.

However, the use of this approach requires zero boundary conditions necessary for accurate undistorted reconstruction of the desired image of the subject. In practice, these zero boundary conditions can be ensured by placing an opaque screen with a hole of a given shape and size on the path of the scanning beam, which will allow irradiating only the necessary area of the object under study, ensuring that there is no radiation outside this region of the subject (zero boundary conditions). In this case, the width of the scanning beam can be significantly larger than the size of the hole in the screen.

In this case, it is also possible to restore an unchanged image if the distribution of the radiation intensity in a wide scanning beam (larger than the studied area) is known in advance, and this distribution is inhomogeneous (uneven) over the beam cross section. The requirement of heterogeneity over the beam cross section in this case is obvious, because if it is not fulfilled, then this leads to the effect of simple illumination without any scanning. Recovery is based on the reverse convolution operation.

Further, this approach can be further generalized by using such an irradiation of an object with a wide beam, which will lead to a result equivalent to the cyclic convolution of the desired image and the inhomogeneous distribution of the radiation intensity in the beam cross section.

This can be achieved by irradiating an object with a wide beam with a given inhomogeneous intensity distribution in its cross section, and then changing the intensity distribution in the beam cross section in time by cyclic shift of the intensity distribution in the beam cross section in the direction corresponding to the scanning direction.

In this case, the scan result will be characterized by changes described by the cyclic convolution of the desired image and the distribution of the radiation intensity in the cross section of the scanning beam. Consequently, the desired image can be restored based on the inverse cyclic convolution, knowing the intensity distribution in the cross section of the scanning beam.

The proposed technical solutions are based on this approach. The intensity distribution in the cross section of the wide scanning beam is set in advance, and in such a way as to maximize the accuracy of the final result under noise and measurement errors.

In the proposed technical solutions, each parameter of the subject is not assigned a certain parameter (for example, frequency), on the basis of which the separation of signals coming from different parts of the subject could occur. There is no preliminary splitting of an object into parts or its equivalent exposure to multiple rays.

The proposed technical solutions are quite simply implemented both when using a fan beam and when using a conical beam, and allow you to effectively use the power of the source of penetrating radiation.

These technical solutions can be implemented, for example, on the basis of a screen inhomogeneous for penetrating radiation (consisting of a cyclically repeating pattern, each cycle of which is not less than the width of the irradiating beam), which moves in the beam section. If the screen moves in the direction of repetition of the pattern, then the signal measured by the detector will be the result of the cyclic convolution of the desired image and the given pattern. Knowing it, you can restore the desired image. Under conditions of noise and measurement errors, the pattern is selected in such a way as to maximize the accuracy of the final result.

For the one-dimensional case (fan beam), the pattern should be repeated along one coordinate, for the two-dimensional case (conical beam), the pattern should be repeated in two coordinates.

If the pattern can be represented as the product of two patterns that are independent in each of the coordinates, then the convolution operation can be performed independently in each of the coordinates by moving each of the patterns in its own coordinate. Recovery is performed sequentially based on the reverse convolution operation, first for one coordinate and then for another. Such actions can reduce the required number of calculations and simplify the device.

However, this approach to improving the efficiency of using a source of penetrating radiation is also not the only possible one. Classical scanning can be extended based on the well-known Radon transform [6]. The essence of this extension of the classic scan is as follows.

Suppose we want to increase the efficiency of using a source of penetrating radiation by expanding a narrowly focused scanning beam. However, we expand the beam in only one direction (to the fan), and to such a width that it becomes wider than the subject under study. The signal registered by the detector in this case will represent the integral along the illuminated direction [6].

Next, we will scan with this beam, getting the values of the integrals along the illuminated directions, and we will keep the scanning direction perpendicular to the plane of the beam.

We will scan in one direction in the image field of the object, then turn the beam (or object) and scan in the other direction, then in the third, and so on, until we get a full scan of 180 degrees. Having a complete set of data, you can restore the image of the studied subject with high spatial resolution [6].

Such a procedure is widely used in computed tomography, and mathematically described on the basis of direct and inverse Radon transforms [6]. The described variant in tomography corresponds to the so-called "parallel geometry", however, the path of the beam can be chosen so that it would lead to the fixation of one edge of the beam outside the object and the rotation of this beam around this point. The data thus obtained will be in the so-called "fan geometry" [6], for which algorithms for reconstructing the desired image [6] have also been developed. Full scan for "fan geometry" is a 360 degree scan. The terms “parallel” and “fan” refer to computed tomography [6], where they are well-established.

Therefore, in this case, modifying the classical scan by changing the size of the beam and the scanning path, you can get an undistorted image of the studied object and increase the efficiency of use of the radiation source.

However, in this case, only a part of the conical beam of the source of penetrating radiation collimated to a flat fan is used. I would like to use a fully conical beam. This problem can be solved by combining the first and second approaches, that is, use the convolution and the Radon transform at the same time. In the case of combining, scanning should be performed as in the Radon transform (parallel or fan geometry), but instead of a flat beam, a beam expanded in the scanning direction should be used. Moreover, the intensity distribution in the beam in the direction of expansion should be organized, as for a one-dimensional version of convolution. In this case, after the scan is completed, before performing the inverse Radon transform, the reverse convolution operation should be performed on the basis of knowledge of information about a given distribution of radiation intensity in the indicated direction.

If you use a sufficiently large beam expansion in the scanning direction, then you can get the full use of the conical beam. In this case, the required resolution in the direction of beam expansion can be restored on the basis of the reverse convolution operation.

Thus, by modifying the classical scanning based on the use of a narrowly collimated beam in various ways, it is possible to increase the efficiency of using a penetrating radiation source without using a coding approach based on assigning a certain parameter (for example, frequency) to parts of an object.

As follows from the above, the fundamental feature of the proposed inventions is to use a wide beam in the study of the subject, rather than a narrowly collimated beam, and at the same time to obtain high spatial resolution and a high signal to noise ratio in the resulting image. The signal C (x, y) recorded by the detector from the object in this case is the result of the convolution of the desired image of the studied object I (x, y) and the radiation intensity distribution function in the beam cross section H (x, y).

For a fan beam, that is, a one-dimensional case, this corresponds to

where s is the integration variable;

I (x) - the desired image of the studied subject (that is, the image scanned by a needle with an infinitely narrow beam);

H (x) is the distribution of the radiation intensity in the cross section of the fan beam (ie, the intensity distribution in the "spot" irradiating the object);

C (x) is the signal recorded by the detector.

Moreover, taking into account the fact that the beam moves (scans)

where V is the scanning speed;

t is the scan time;

the signal recorded by the detector is a function of time t.

For a conical beam, that is, a two-dimensional case,

where η is the integration variable;

I (x, y) - the desired image of the studied subject;

H (x, y) is the distribution of the radiation intensity in the section of the conical beam;

C (x, y) is the signal recorded by the detector.

Subsequent execution of the reverse convolution (deconvolution) operation on the signal recorded by the detector allows you to restore the undistorted desired image of the object under study.

For the one-dimensional case

where H ^-1 (x) is the restoring function for H (x).

For the two-dimensional case

where H ^-1 (x, y) is the restoring function for H (x, y).

If the function H (x, y) of the distribution of the radiation intensity in the beam section for the two-dimensional case can be written as the product of two functions H ₁ (x) and H ₂ (y) along each of the coordinates:

then the operations of convolution and deconvolution can be carried out separately for each of the coordinates:

which simplifies the implementation, as it allows you to conduct operations separately for each of the coordinates.

If the function H (x} or H (x, y) is a delta function, which corresponds to scanning an object with a highly directed beam, then restoration is not required, since the signal recorded by the detector exactly corresponds to the image of the studied object.

If the function H (x} or H {x, y) is a constant, which corresponds to the complete absence of scanning and the illumination of the object simply by a wide uniform beam, then restoration is generally impossible.

Thus, the fundamental requirement for the distribution function of the radiation intensity in the beam cross section is the requirement that the function differs from the constant. If this requirement is met, then it is possible to carry out a direct and inverse transformation with varying degrees of certainty. From a mathematical point of view, this problem belongs to the so-called class of incorrect inverse problems, the solution of which was developed by academician A.N. Tikhonov [4] and is based on regularization. The essence of these decisions based on regularization is most easily stated using the frequency approach.

If F [C (x, y)] is the Fourier transform of the recorded signal, and F [H (x, y)] is the Fourier transform of the intensity distribution in the section of the scanning beam, then the Fourier transform of the desired image is F [I (x, y)] of the investigated subject is equal

where α is the regularization coefficient.

The image itself is equal

where F ^-1 [...] is the inverse frequency conversion operator.

That is, the Fourier transform of the recovery function F [H ^-1 (x, y)] is equal to:

and the recovery function itself is equal to

Consequently, the essence of the coefficient of regularization a consists in “protection” from dividing by small quantities. Its specific value is selected depending on the nature of the function F [H (x, y)].

For the one-dimensional case (fan beam), the relations are similar.

The above description of the essence of regularization is a particular one and for a more detailed explanation, refer to [4].

In a practical implementation using digital computer technology, the signal recorded by the detector is sampled in time. Continuous integration is converted to a summation of discretized quantities.

For a fan beam (one-dimensional case), the signal recorded by the detector is

and the desired discrete image is equal to

For a conical beam (two-dimensional case), the signal recorded by the detector is

and the desired discrete image is equal to

where N and M are determined by the real length of the arrays H (...), I (...) and C (...)

For the distribution function of the radiation intensity in the beam section, which is the product of two functions,

which reduces the number of calculations.

In this case, the discrete recovery functions H ^-1 (...) can be found based on the knowledge of the discrete distributions of the radiation intensity in the cross section of the scanning beam H (...) and the regularization approach described earlier [4], together with the discrete Fourier transform.

However, in addition to using aperiodic convolution, it is possible to use cyclic convolution. In this case, the data can be presented arranged cyclically with a repetition period M for a one-dimensional case or with repetition periods N and M for a two-dimensional case.

In the case of using cyclic convolution, the relations take on a slightly different form.

For a fan beam (one-dimensional case)

For a conical beam (two-dimensional case)

For the distribution function of the radiation intensity in the beam section, which is the product of two functions,

which reduces the number of calculations.

In the above relations, mod [..., N] or mod [..., M] means the calculation of the value modulo N or M of the specified value.

In this case, for both cases (one-dimensional or two-dimensional, aperiodic or cyclic convolution), the higher the sampling rate of the signal measured by the detector, the greater the spatial resolution of the image of the object under study.

That is, with this approach, as with classical scanning, the spatial resolution (to a first approximation) is determined by the sampling frequency of the signal measured by the detector. If the distribution function of the radiation intensity in the beam cross section is known, then to obtain a higher resolution image, you just need to increase the sampling frequency of the signal without changing the distribution of the radiation intensity in the beam cross section.

In the presence of noise, which corresponds to a real practical situation, the choice of an arbitrary convolution function (an arbitrary distribution of the radiation intensity in the beam cross section) will not be a practical solution to the problem. A more practical option would be to choose a convolution function that, for a given sampling frequency, maximizes the signal-to-noise ratio in the image of the object under study by some criterion, for example, by the criterion of the maximum signal-to-noise ratio for each individual image element. In this case, the solution of the problem in discrete form is known based on the use of arrays having flat side lobes in cyclic autocorrelation functions, the so-called URA arrays [5]. These arrays are based on the use of Walsh functions, that is, on the basis of 0 and 1. For these arrays, the restoring functions H ^-1 (...) [5] are known, which will consist of -1 and +1. In addition, it is known that the maximum signal-to-noise ratio in each individual element of the reconstructed image is achieved for URA arrays having an effective transparency with respect to penetrating radiation of 50% (i.e., the number of units is equal to 50% of the total array length). Therefore, up to 50% of the power of the penetrating radiation can be effectively used. Using URA arrays with a different transparency coefficient (greater than or less than 50%) does not interfere with operability, but lowers the signal-to-noise ratio in the resulting image.

Making heterogeneous screens based on URA arrays is also very simple, since it comes down to making holes in the screen.

Using other optimization criteria may lead to the use of other convolution functions, but will not change the method of solving the problem.

Thus, the global conclusion that follows from the proposed consideration is that with the proper choice of the convolution function (distribution of the radiation intensity in the beam cross section), it is possible not only to obtain high spatial resolution, but also to efficiently use the power of the penetrating radiation source.

At the same time, if we do not consider the problem of maximally utilizing the power of the source of penetrating radiation, then practically any function can be used as a convolution function (distribution of the radiation intensity in the beam cross section) even under noise conditions. In this case, the final resolution in the resulting image will be determined only by the sampling frequency of the measured data, and the form of the convolution function will affect the signal-to-noise ratio in the desired image. In this case, the image restoration problem can be solved on the basis of discrete versions of the previously mentioned regularization methods proposed by academician A.N. Tikhonov [4] (in the time or frequency domain).

Thus, from what has been said, it becomes clear that the proposed technical solutions differ from those known from the application for the invention [3], where each image element is assigned a certain parameter, for example, frequency.

At the same time, for the two-dimensional case, the proposed approach can be generalized somewhat and instead of convolution (or together with convolution), use the direct Radon transform [6], and when reconstructing the image, use the inverse Radon transform [6].

This transformation is widely used in transaxial computed tomography and is reduced to obtaining a set of projections of the studied section at different angles:

where the x ’and y’ axes are set by turning it counterclockwise

Subsequent restoration of the image of the studied section on the basis of the measured set of projections, that is, obtaining the inverse R ^-1 Radon transform [6], in the operator form can be written as

where D _y is the partial derivative operator with respect to the first variable;

H _y is the Hilbert transform operator;

B is the back projection operator.

For a more detailed description of the direct and inverse Radon transforms, one should turn to the source indicated earlier [6]. There you can find several discrete options for the implementation of the inverse Radon transform for fan and parallel geometry, usually called reconstruction algorithms [6] and used in digital computers.

In the proposed technical solutions, this transformation is used to increase the efficiency of using the source of penetrating radiation and, on the basis of this, to accelerate scanning processes. The increase in efficiency is achieved through the use of a wide scanning beam, which realizes taking the integral along a given direction in the Radon transform. Thus, per unit time, an object is irradiated with more quanta than when scanning with a narrow beam.

The invention is illustrated by drawings, on which:

figure 1 - transformation of classical scanning to scanning using a wide conical beam and cyclic convolution;

figure 2 - heterogeneous screen, consisting of four repeating parts, designed to perform two-dimensional cyclic convolution;

figure 3 - separation of the inhomogeneous screen into two parts to perform two-dimensional cyclic convolution independently in two coordinates;

figure 4 - transformation of the classical scan to scan using a wide flat beam and the Radon transform;

figure 5 - system for scanning with a fan beam;

figure 6 - system for scanning a conical beam;

7 is a two-dimensional scan based on a one-dimensional cyclic convolution using a drum rotating in one direction;

on Fig - scanning based on the implementation of two-dimensional cyclic convolution by moving a non-uniform screen;

figure 9 is a scan based on a two-dimensional cyclic convolution using a single drum moving in two directions;

figure 10 is a scan based on a two-dimensional cyclic convolution using one sphere moving in two directions;

figure 11 is a scan based on the Radon transform for "parallel geometry";

on Fig - scan based on the Radon transform for "parallel geometry" and one-dimensional cyclic convolution;

on Fig - scan based on the Radon transform for "fan geometry";

on Fig - scan based on the Radon transform for "fan geometry" and one-dimensional cyclic convolution;

on Fig - scanning based on a two-dimensional cyclic convolution independently in two coordinates by moving two screens;

on Fig - scanning based on a two-dimensional cyclic convolution independently in two coordinates using a drum;

on Fig - scanning based on a two-dimensional cyclic convolution independently in two coordinates using two reels;

on Fig is another scan option based on a two-dimensional cyclic convolution independently in two coordinates using two drums;

on Fig - simulation results.

The transformation of classical scanning to scanning using a wide conical beam and cyclic convolution is shown in figure 1.

On figa presents a classic scan using a narrowly collimated beam of penetrating radiation. The formed narrow-collimated beam 9a from the source 7a performs two-coordinate movement (for example, line-by-line), scanning the object 6.

Classical scanning is performed with a narrow beam, when the beam position fully characterizes the position of the part of the object being irradiated. In this case, it is not necessary to perform any coding of information coming from the part of the object being irradiated at the moment, since the position of the beam completely determines the part of the object being irradiated at the moment.

A narrowly collimated beam 7a is usually formed by collimating a wider beam emitted by a source of penetrating radiation. In this case, a significant part of the initial beam power is lost.

In order to increase the efficiency of using the power of the penetrating radiation source, it is possible to expand the beam width — see FIG. 1B, where 7b denotes the source forming the expanded beam 9b.

If you scan an object with an expanded beam 9b along the same trajectory as in figa, this will lead to the fact that in a unit time the object will be irradiated with a large number of quanta. However, the scan result will be changed. The nature of the changes will be characterized by the convolution of the desired image (scanned by a narrow beam) and the distribution of the radiation intensity in the beam section 9b. Nevertheless, if the distribution of the radiation intensity in the section of the beam 9b is known, then it is possible to perform the reverse convolution and restore the undistorted desired image.

Further, this approach can be further generalized, based on the classical scanning of an investigated object with an expanded beam, expanding the beam to sizes equal to or larger than the studied object - see Fig. 1C, where 7c denotes a source of penetrating radiation that forms a very wide beam 9c (i.e. practically without additional collimation of the initial radiation). If you scan with such a beam 9c along the same trajectory as in figa, the object per unit time will be irradiated with even more quanta. The scan result will be different from FIG. However, the nature of the changes in the desired image will not change, and will also be characterized by the convolution of the desired image and the radiation intensity distribution in the cross section of a very wide scanning beam 9c. Consequently, the unchanged desired image can be restored by performing reverse convolution if the distribution of the radiation intensity in such a very wide scanning beam is known.

However, the use of this approach requires zero boundary conditions [4], which are necessary for carrying out an accurate undistorted restoration of the desired image of the object. In practice, these zero boundary conditions can be ensured (see FIG. 1D) by placing an opaque screen with an opening of a given shape and size 8 on the path of the scanning beam 9d, which will allow irradiating only the necessary studied region of object 6, ensuring that outside this region subject no exposure (zero boundary conditions). Moreover, the width of the scanning beam 9d can be significantly larger than the size of the hole in the screen and the radiation 10d generated by it.

Scanning should be done along the same path as in figs. The scan result will be different from FIG. The nature of the changes will continue to be characterized by the convolution of the desired image and the radiation distribution in the beam section 9d.

It is also possible to restore an unchanged image if the distribution of the radiation intensity in a wide scanning beam 9d (larger than the studied area) is known in advance, and this distribution is inhomogeneous (uneven) over the beam cross section. The requirement of inhomogeneity over the beam cross section in this case is obvious, because if this is not fulfilled, then this leads to the effect of simple illumination without any scanning. Recovery is based on the reverse convolution operation [4].

However, to use scanning with a very wide beam is inefficient, since only a small part of it interacts with the object. I would like the whole wide beam to constantly interact with the object.

To this end, this approach can be further generalized using such irradiation of the subject 6 or its predetermined part (see FIG. 1E) with a wide beam 10e formed by the collimator 8, which will lead to a result equivalent to the cyclic convolution of the desired image or its predetermined part and heterogeneous distribution of radiation intensity in the beam cross section 10e. In this case, an inhomogeneous intensity distribution in its cross section can be formed by a screen 5 placed on the beam propagation path 10e and containing an inhomogeneity cyclically repeated along two coordinates in the form of sections 1, 2, 3, and 4 (see Fig. 2). In this case, the dimensions of the inhomogeneity 1 must be not less than the beam width 10e. Scanning should be carried out along the same path as in fig.1D, by moving in time the screen 5 along this path.

In this case, the scan result will be characterized by the changes described by the cyclic convolution of the desired image and the distribution of the radiation intensity in the cross section of the scanning beam (in accordance with the inhomogeneity 1 of screen 5). Knowing the distribution of the radiation intensity in the section of the scanning beam 10e (knowing the distribution of the inhomogeneity of the portion 1 of the screen 5), it is possible to reconstruct the desired image based on the inverse cyclic convolution.

The proposed technical solutions are based on this approach, with the intensity distribution in the cross section of the wide scanning beam (distribution of the inhomogeneity of 1 screen 5) set in advance, so that in the conditions of noise and measurement errors, maximize the accuracy of the final result.

If the distribution of heterogeneity 1 of screen 5 can be represented separately as the product of heterogeneities for each of the coordinates, then screen 5 can be divided into two screens 5a and 5b (see FIG. 3) containing cyclically repeating inhomogeneities 1a and 1b along each of the coordinates. In this case, instead of scanning based on the two-coordinate movement of the screen 5, scanning based on the single-coordinate movement of each of the screens 5a and 5b (Fig. 3) should be used, which can greatly simplify the implementation of this scanning method.

In the proposed technical solutions, each parameter of the subject is not assigned a certain parameter (for example, frequency), based on which the separation of signals coming from different parts of the subject occurs. There is no preliminary partitioning of an object into parts, which is equivalent to irradiation with a multitude of rays used in the closest technical solutions [3].

However, this approach to improving the efficiency of using a source of penetrating radiation is also not the only possible one. The classical scanning can be extended with a narrowly collimated beam using the Radon transform [6]. The essence of this extension of the classic scan will be as follows.

Suppose we want to increase the efficiency of using a source of penetrating radiation by expanding a narrowly focused scanning beam 9a formed by a source 7a (see Fig. 4A). However, we expand the beam 9a in only one direction (to the fan-shaped), and to such a width that it becomes wider than the object under study (position 9f in Fig. 4B). The signal registered by the detector in this case will represent the integral along the direction illuminated on the object 6.

Next, we will scan with this beam, getting the values of the integrals along the illuminated directions, and we will keep the scanning direction perpendicular to the plane of the beam. The signal S (t, θ _i ) recorded by the detector in time will represent the values of the integrals along the scanning direction [6].

We will scan in one direction in the image field of an object for angle θ ₁ , then turn the beam (or object) and scan in another direction for angle θ ₂ , then in the third for angle θ ₃ , and so on, until we get a full scan on 180 degrees - see figs. Having a complete data set S (t, θ _i ), it is possible to reconstruct the image of the studied object with high spatial resolution [6].

The indicated restoration procedure is widely used in computed tomography, and is mathematically described on the basis of the inverse Radon transform [6]. The described version in tomography corresponds to the so-called "parallel geometry", however, the beam path can be chosen so that it would seem to fix one edge of the beam 9f outside the object and rotate this edge of the beam 9f outside the object and rotate this beam around this point - see fig.4D. The data thus obtained (see FIG. 4E) will be in the so-called "fan geometry" [6], for which algorithms have also been developed for reconstructing the desired image [6] from the data set S (t, θ _i ). Full scan for "fan geometry" is a 360 degree scan. The terms “parallel” and “fan” refer to computed tomography [6], where they are well-established.

Therefore, in this case, modifying the classical scan by changing the size of the beam and the scanning path, you can get an undistorted image of the studied object and increase the efficiency of use of the radiation source.

However, in this case, only a part of the conical beam of the source of penetrating radiation collimated to a flat fan is used. I would like to use a fully conical beam. This problem can be solved by combining the first and second approaches, that is, to use convolution and the Radon transform at the same time. In the case of combining, scanning should be performed as in the Radon transform (parallel or fan geometry), but instead of a plane beam, a plane beam expanded in the scanning direction should be used. Moreover, the intensity distribution in the beam in the direction of expansion should be organized as for a one-dimensional version of the cyclic convolution. In this case, after the scan is completed, before performing the inverse Radon transform, an inverse cyclic convolution operation should be performed on the basis of knowledge of information about a given distribution of radiation intensity in the indicated direction.

If you use a sufficiently large beam expansion in the scanning direction, then you can get the full use of the conical beam. In this case, the required resolution in the direction of beam expansion can be restored on the basis of the reverse convolution operation.

To scan a given part of the subject and create zero boundary conditions necessary for the recovery operation, you should use the collimator 8 (not shown in Fig. 4), forming a beam of a given size. Scanning itself can be carried out, as before, based on the displacement of the inhomogeneous screen 5 in the beam of penetrating radiation in a given direction - see Fig. 1E. However, instead of the screen 5 having a heterogeneity cyclically repeated in two coordinates, one should use a screen 5a with a heterogeneity 1a repeating in one direction (see Fig. 3), which will ensure that a cyclic convolution along a given direction is performed.

Thus, by modifying the classical scanning based on the use of a narrowly collimated beam in various ways, it is possible to increase the efficiency of using a penetrating radiation source without using a coding approach based on assigning a certain parameter (for example, frequency) to parts of an object [3].

Devices for implementing the proposed method are shown in Fig.5 and Fig.6. Figure 5 shows a device for scanning with a fan beam (one-dimensional case), figure 6 - device for scanning with a conical beam (two-dimensional case). They are distinguished by the presence of a conveyor 36 for moving the test object 6 past the scanning beam 10. The use of a conveyor 36 for scanning with a conical beam is not required, since scanning is carried out in two coordinates, in addition, as has been repeatedly noted, the use of a conical beam is more advantageous from an energy point of view.

Therefore, we consider in detail the device shown in FIG. 6, considering that the device shown in FIG. 5 is a particular implementation thereof.

6, reference numeral 30 shows a scanner that implements a time-varying radiation intensity in a section of a conical scanning beam 10. An object 6 is mounted on the path of the scanning beam of penetrating radiation 10, followed by a transmitted radiation detector 32 (segmented or non-segmented) in the path of the beam 10 and, outside the beam 10, a scattered radiation detector 31 is installed (segmented or not). The scattered radiation detector can be installed as a detector of frontal scattered radiation, side scattered radiation and as a detector of rear scattered radiation. 6, detector 31 is shown as a front scattered radiation detector. Both detectors 31 and 32 are connected to the processor 34, which is also connected to the visualization tool 33 and the control unit 35, which is connected to the scanner 30 and the detectors 31 and 32,

The purpose of the detectors 31 and 32 is the registration of the intensity of the penetrating radiation in time after interaction with the object 6.

The device shown in Fig.6 operates as follows.

The scanner 30 forms a conical beam 10 of penetrating radiation with a time-varying distribution of the radiation intensity in its section, in accordance with the selected scanning method. The beam of penetrating radiation 10 interacts with the object 6. After interacting with the object, the intensity of the penetrating radiation is measured by the detectors 31 and / or 32. The measurement results in time from the output of the detectors 31 and / or 32 are transmitted to the processor 34, where they are stored to the extent necessary for restoring on their basis an undistorted image of the investigated subject.

In this case, the detectors 31 and / or 32 may contain signal preamps and analog-to-digital converters (ADCs) and transmit the signal to the processor already in digital form.

After the signals from the detectors 31 and / or 32 are memorized to the extent necessary for the calculations, they are used to make calculations to obtain an undistorted image of the object under study. The result of the calculation can be visualized on the visualization tool 33 or transmitted via communication channels to an external consumer.

In this case, before the start of scanning and after it is finished, signals are sent from the computer 34 or manually through the control unit 35 from the computer 34 or allow setting the scanning parameters for the scanner 30 and the detectors 31 and / or 32. These signals for the scanner may include signals that establish the power of the penetrating radiation source , the energy range of the source, scanning speed, the beginning of on and off, and others. The detectors 31 and / or 32 from the control unit 35 can be fed the values of the gain for the preamplifiers, sampling interval and others.

The mention that the signals in the processor 34 are remembered to the extent necessary to restore an undistorted image of the subject under study on their basis should be understood as the possibility of starting processing before the end of the full scanning process. This can be useful to reduce data processing time. In this case, the initial data itself can be fully saved for the entire scanning process for subsequent use.

The difference in the operation of the devices shown in FIG. 5 and FIG. 6 is that in the device of FIG. 5, the object 6 during scanning is moving on the conveyor 36 in a direction perpendicular to the plane of the beam 10.

The scanner 30 in the conical beam scanning device of FIG. 6 can be performed in various ways depending on the selected scanning method.

Figure 7 presents the scanner, made on the basis of a single screen inhomogeneous for penetrating radiation, moving in one direction. Moreover, the moving inhomogeneous screen is a drum 25, rotating around its axis and having a one-dimensional pattern of length M × N, repeated along the height of the drum N times, with a shift by a length M for each subsequent repeat in height. One revolution of the drum corresponds to a full two-dimensional scanning based on a cyclic one-dimensional convolution.

Theoretically, this version of the scanner is somewhat different in scanning methods based on cyclic convolution, but it has the simplest implementation.

The drum 25 is driven by an engine 15, providing uniform rotation of the drum 25 at a given speed. In this case, the source of penetrating radiation 7 is located in the center of rotation of the drum 25. For the formation of a beam 10 of a given size, the source 7 may be provided with a collimator (not shown) located inside the drum 25. However, the collimator may be located behind the drum 25, forming a beam 10 of a given size. In this sense, the position of the collimator is not critical.

The undistorted image of the scanned object is restored in this version of the scanner on the basis of a one-dimensional inverse cyclic convolution of the length M × N of elements taking into account the known distribution of heterogeneity in the pattern of length M × N located on the walls of the drum.

After performing reverse convolution, the result will be a one-dimensional array of numbers of length M × N, where the first M numbers will characterize the first line of the image, the second M numbers will characterize the second line of the image, and so on. The total number of lines in the restored image will be equal to N.

If a pattern of length M × N is cyclically arranged around the drum circumference with the number of cycles K, then a full scan will be completed in a time corresponding to more than one revolution of the drum 25, but K times smaller.

In addition to the implementation of the scanner on the basis of one inhomogeneous screen moving in one direction, it is possible to implement it on the basis of one inhomogeneous screen moving in two directions. The simplest version of such a scanner is shown in Fig. 8. In this simplest case, the non-uniform screen 5 is a plate mounted on a two-axis linear motor 11. To generate a beam 10 of a given size, the penetrating radiation source 7 is provided with a collimator 8 installed in front of the plate (screen) 5. In this case, the plate Between the plate (screen) 5 and the source of penetrating radiation 7 contains a cyclically repeating heterogeneity 1 in accordance with figure 2.

The operation of this scanner consists in moving the plate (screen) 5 at first in one direction, for example, horizontal (which corresponds to scanning the image in rows). Then follows the movement of the screen 5 up (or down) by a predetermined distance and the next scan in the horizontal direction in the opposite direction follows. Scanning ends when all the lines of the desired image are traversed.

However, the operation of a two-axis linear motor in a reciprocating mode may not be fast enough. Therefore, FIG. 9 shows a variation of the scanner where there is no reciprocating movement, but the inhomogeneous screen also moves in two directions.

Figure 9 presents the scanner, made on the basis of a single inhomogeneous for penetrating radiation screen moving in two directions. In this case, the moving inhomogeneous screen is a drum 19, rotating around its axis and moving along its axis, with a two-dimensional pattern deposited on its surface, cyclically repeating along the height and circumference of the drum.

The drum 19 is driven by an engine 15, providing uniform rotation of the drum 19 with a given speed. While the source 7 of penetrating radiation is located in the center of rotation of the drum 19.

To form a beam 10 of a given size, the source 7 may be provided with a collimator (not shown) located inside the drum 19 (or behind the drum 19).

The engine 15 is connected to a linear motor 14, providing movement of the drum 19 in the direction of the axis of rotation of the drum 19.

The rotation of the drum provides scanning along one coordinate (for example, in the direction of image lines), and linear movement along another (for example, in the direction of image columns). If one two-dimensional pattern is located around the circumference of the drum 19, then scanning on one line ends after a full revolution of the drum. Then it should be raised using the engine 14 to a predetermined amount and scan the next line, and so on. A full scan ends after N revolutions of the drum 19, where N is the number of lines in the desired image.

If a two-dimensional pattern 1 is arranged around the drum circumference cyclically with the number of cycles K, then a full scan will be completed in a time corresponding to N / K revolutions of the drum 19.

At high penetrating radiation energies, the wall thickness of the drum 19 will have to be made thick enough to ensure effective interaction of the penetrating radiation of the source 7 with the material of the drum 19. Moreover, the interaction (absorption) of radiation from the source 7 with the material of the drum 19 will depend on the angle at which the beam falls to the surface of the drum. It is more rational to place the pattern not on the surface of the drum, but on the surface of the sphere. In this case, the radiation from the source 7, located in the center of rotation of the sphere, will uniformly interact with the drum material (absorbed) regardless of the position of the beam in the beam.

Figure 10 presents the scanner, made on the basis of a single inhomogeneous for penetrating radiation screen moving in two directions. In this case, a screen moving in two directions, inhomogeneous for penetrating radiation, is a sphere (ball belt) 20, rotating simultaneously around two orthogonal axes passing through its center, and having a two-dimensional pattern on its surface, applied cyclically relative to the indicated rotation axes.

The sphere 20 is set in motion by the engine 15, ensuring its uniform rotation with a given speed. In this case, the source of penetrating radiation 7 is located in the center of rotation of the sphere 20. To form a beam 10 of a given size, the source 7 can be equipped with a collimator (not shown) located inside the sphere 20 (or behind the sphere 20).

The engine 15 is connected to the engine 17 through a rod 22, one end attached to the engine housing 15, and the other to the axis of the engine 17 so that the sphere 20 as a result of the operation of the engines 15 and 17 rotates around two orthogonal axes passing through its center.

Rotation of a sphere around one axis provides scanning along one coordinate (for example, in the direction of image lines), and rotation around another axis - scanning along another coordinate (for example, in the direction of image columns). If one two-dimensional pattern is located around the circumference of the sphere 20, then scanning on one line ends after the sphere 20 is completely rotated from the engine 15. Then it should be turned with the help of the engine 17 to a predetermined angle and scanning of the next line, and so on. A full scan ends after N revolutions of the sphere 20 from the engine 15, where N is the number of lines in the desired image.

If we arrange a two-dimensional pattern 1 around the circumference of sphere 20 cyclically with the number of cycles K, then a full scan will be completed in a time corresponding to N / K revolutions of sphere 20.

As indicated earlier, scanning can be carried out not only on the basis of convolution, but also on the basis of the Radon transform for "parallel geometry".

A scanner containing one inhomogeneous screen for penetrating radiation, moving in two directions and scanning based on the Radon transform for "parallel geometry", is shown in Fig. 11. In this scanner, a screen inhomogeneous for penetrating radiation, moving in two directions, is a drum 21 rotating around its axis, and the axis of the drum rotates around an axis orthogonal to it. On the surface of the drum 21, a one-dimensional pattern (slots) is applied around the circumference. The rotation of the drum 21 around its axis is provided by the engine 15, and the rotation around the orthogonal axis by the engine 17, the axis of which is connected through the rod 22 to the housing of the engine 15. The source of penetrating radiation 7 is located in the center of the drum 21 on its axial line. Between the source 7 and the drum 21 is a collimator (not shown), which provides the formation of a beam 10 of a given size.

If the drum 21 has one slot, then obtaining one projection corresponds to one revolution of the drum 21 from the engine 15. To obtain another projection, rotate the drum 21 at a predetermined angle using the engine 17 and get another projection, and so on. A full scan ends after engine 17 rotates 180 degrees.

If the drum 21 has K slots, then obtaining one projection during rotation of the engine 15 is reduced by a factor of K. However, obtaining a full scan requires an engine rotation of 17 by 180 degrees.

A scanner combining scanning based on the Radon transform for "parallel geometry" and convolution is shown in Fig. 12. This scanner differs from the scanner shown in Fig. 11 in the form of a rotating drum. In this drum 23, instead of slots located through a given angle, there is a one-dimensional inhomogeneous pattern 1a located cyclically through the same angle. Otherwise, the scanners of FIG. 11 and FIG. 12 are the same. Their work coincides, but the nature of the data processing is different. Before calculating the desired image of the scanned object on the basis of the inverse Radon transform for "parallel geometry", it is necessary to perform the inverse cyclic convolution for each projection based on the knowledge of the distribution of inhomogeneity 1a.

As indicated earlier, scanning can be carried out not only on the basis of the Radon transform for "parallel geometry", but also on the basis of the Radon transform for "fan geometry".

A scanner containing one inhomogeneous screen for penetrating radiation, moving in two directions and scanning based on the Radon transform for "fan geometry", is shown in Fig. 13. In this scanner, a screen moving in two directions, inhomogeneous for penetrating radiation, is a disk 24, rotating around its center, and the center of the disk rotates in a circle lying in its plane (planetary motion), with a one-dimensional pattern (slots) applied to the surface of the disk ) relative to the center of rotation of the disk.

The rotation of the disk 24 around its axis is provided by the engine 17a, and the planetary movement is provided by the engine 17b, the axis of which is connected through the rod 22 to the housing of the engine 17a. In this case, the source of penetrating radiation 7 is located on the line, which is a virtual continuation of the axis of the engine 17b.

To form a beam 10 of a given size, the source 7 is equipped with a collimator 8 located in front of the disk 24.

If the disk 24 has one slot, then obtaining one projection corresponds to one revolution of the disk 24 from the engine 17a. To get a different projection, you should rotate the center of rotation of the disk 24 by a predetermined angle using the engine 17b and get a different projection, and so on. A full scan ends after engine 17b completes one full revolution.

If the disk 24 has K slots, then obtaining one projection during rotation of the engine 17a is reduced by a factor of K. However, obtaining a full scan requires, as before, a complete revolution of the engine 17b.

A scanner combining scanning based on the Radon transform for "fan geometry" and convolution is shown in Fig. 14. This scanner differs from the scanner shown in Fig. 13 in the form of a rotating disk 25. In this disk, instead of slots located through a given angle, there is a one-dimensional inhomogeneous pattern 1a located cyclically through the same angle. The rest of the scanner devices are the same. Their work coincides, but the nature of the data processing is different. Before calculating the desired image of the scanned object on the basis of the inverse Radon transform for "fan geometry", it is necessary to perform the inverse cyclic convolution for each projection based on the knowledge of the distribution of inhomogeneity 1a.

In the case when the distribution of the radiation intensity in the beam cross section can be represented as the product of two functions, as noted earlier, the scan can be organized on the basis of two inhomogeneous screens, each moving in its own direction and implementing the cyclic convolution operation for its direction. In this case, from the Radon transform, we again return to scanning based on cyclic convolution.

The simplest scanner based on two heterogeneous screens moving in different directions is shown in Fig. 15.

The scanner in this case contains a source of penetrating radiation 7 and two plates (screens) 5a and 5b mounted in series on the path of penetrating radiation from source 7 to single-axis engines 12 and 14. Screens 5a and 5b are organized in accordance with Fig.3. To form a beam 10 of a given size, the source 7 is equipped with a collimator 8 mounted in front of the screen 5a.

The scanner consists in linearly moving the screens 5a and 5b along orthogonal coordinates using linear motors 12 and 14. The engine 12 moves the screen 5a by a predetermined distance (equal to the width of the pattern 1a) along the direction of the lines. Then, the engine 14 moves the screen 5b by a predetermined step in the direction corresponding to the columns of the image, after which the engine 12 performs a return movement. This happens until the screen 5b is moved a predetermined distance corresponding to the width of the pattern 1b. This completes the scanning process.

This scanner option is the simplest and may not provide high scanning speeds due to the presence of reciprocating screens. On Fig shows a modification of this scanner, when one of the screens is made in the form of a rotating drum 16. This drum replaces the screen 5A in Fig.15, since it is he who must make the most frequent reciprocating movements. On the surface of the drum, a one-dimensional inhomogeneous pattern 1a is applied around the circumference. If one pattern 1a is located on the circumference of the drum, then scanning in the direction of the rows corresponds to one revolution of the drum 16. If K patterns 1a are located cyclically on the circumference of the drum, the scanning time in the direction of the rows will be reduced by a factor of K.

A further development of this approach is a scanner based on two rotating drums.

A scanner comprising two screens inhomogeneous for penetrating radiation, each moving in its own direction, is shown in FIG. In this scanner, the moving inhomogeneous screens are drums 16 and 18 with cyclically applied one-dimensional patterns 1a cyclically around their circumference. The drums are placed one into the other and rotate around their axes, oriented orthogonally relative to each other.

The source of penetrating radiation 7 is placed at the intersection of the axes of the drums 16 and 18. To form a beam 10 of penetrating radiation of a given size, the source 7 can be equipped with a collimator (not shown) placed inside the drums (or behind them). The drum 16 is driven by the engine 15, and the drum 18 is driven by the engine 17.

The rotation of the drum 16 corresponds to scanning the image in one coordinate, and the rotation of the drum 18 corresponds to scanning in a different coordinate. One revolution of the drum 16 or 18 corresponds to a full scan along its coordinate if there is only one pattern of heterogeneity on the surface of the drums. If K cyclically repeating patterns 1a and 1b are deposited on the surface of the drums, then the scanning time along each coordinate occurs not in one revolution of the drum, but in K times faster.

If the drum 16 is responsible, for example, for scanning along the rows, then the rotation of the drum 18 (responsible for the columns) should be N times slower, where N is the number of lines in the image.

Making a scanner in which the drums are orthogonal to one another is not the only option for implementing this approach.

On Fig shows a scanner in which two inhomogeneous for the penetrating radiation of the screen, each moving in its own direction, are drums 37 and 38 with cyclically deposited on their circumference one-dimensional patterns 1A and 1b. The drums are placed one into the other, rotate around their axes and are oriented coaxially with respect to each other, and the pattern along the circumference of each of the drums is inclined by +45 and -45 degrees relative to the axis of its rotation, respectively.

The source of penetrating radiation 7 is located on the axis of rotation of the drums 37 and 38. To form a beam 10 of a given size, the source 7 can be equipped with a collimator (not shown) located inside the drums or behind them).

In this design, due to the fact that the patterns are tilted in different directions, scanning as a result occurs for orthogonal coordinates (rows and columns of the image). However, the direction of the rows and columns is rotated relative to the design of FIG. 17 by 45 degrees.

The drum 37 is driven by an engine 15 a, and the drum 38 is driven by an engine 15b. If the drum 37 is responsible for scanning, for example, in rows, then the speed of the drum 38 relative to 37 should be N times slower, where N is the number of lines in the image.

If only one heterogeneity 1a or 1b is located around the circumference of the drums 37 and 38, then a full scan in the direction of the row or column is carried out in one revolution of the drum 37 or 38. If K heterogeneities 1a and 1b are located around the circumference of the drums, then the scan is K times faster .

To confirm the operability of the proposed method, a computer simulation of scanning based on cyclic convolution for various types of inhomogeneities 1 was carried out (Fig. 2). The simulation results are presented in Fig.19. On figa shows the image of the investigated object obtained by classical scanning in transmitted radiation. The column of FIG. 19B shows the simulation results when scanning on the basis of a single screen moving in two directions. As heterogeneity 1, a URA array of numbers 211 × 211 elements in size [5] was used, which simulated an inhomogeneous screen made in accordance with the pattern shown in the middle of the column of FIG. 19B. Such a screen can be made, for example, on the basis of photolithography processes, by applying a material opaque to the penetrating radiation onto a transparent substrate.

The top of the column of FIG. 19B shows the result of the cyclic convolution of the image shown in FIG. 19A and the heterogeneity shown in the middle of the column of FIG. 19B (i.e., the scan result). At the bottom of the column of FIG. 19B, a recovery result is shown.

On figs shows similar results for heterogeneity, which can be represented as the product of two functions in independent coordinates, that is, for the case of two screens located one after another, moving independently in different coordinates. As a heterogeneity, a one-dimensional URA array with a length of 255 elements was used [5]. Such screens can be made by milling in a material opaque to penetrating radiation (deposited on a transparent substrate).

On fig.19D shows the simulation results for the heterogeneity, which is negative with respect to the heterogeneity in figs. However, it is not a heterogeneity that can be represented as two screens arranged one after another. However, it can be made of a material opaque to the penetrating radiation (deposited on a transparent substrate) by milling in two orthogonal directions. This is important from a practical point of view. In this case, as in the case shown in FIG. 19B, the movement of one inhomogeneous screen in two directions was simulated.

The screen transparency of FIG. 19B corresponds to 50%, the transparency of the screen of FIG. 19C corresponds to 25%, the transparency of the screen of FIG. 19D corresponds to 75%.

Despite the fact that only imaging based on transmitted radiation was simulated, the simulation results confirm the possibility of using a wide conical beam to inspect objects in both transmitted and scattered radiation, since the nature of the signal and its processing methods for transmitted and scattered radiation do not differ. Moreover, both the optimal function and the non-optimal one, in the sense of increasing the signal-to-noise ratio, can be used as a convolution function. In the conducted mathematical experiments, the signal-to-noise ratio in the signal was quite large, therefore, the resulting images for the optimal and non-optimal functions are practically no different. The differences will be significant only with a very small signal to noise ratio (close to unity). Thus, it is possible to use various options for the technical implementation of the proposed solution in accordance with the available means for their implementation.

Sources of information

1. US patent No. 5,181,234, publ. 01/19/93.

2. US Patent No. 6,546,072, publ. 04/08/2003.

3. US Application No. 20020031202, publ. 03/14/2003.

4. A.N. Tikhonov, V.Ya. Arsenin. Methods for solving incorrect tasks. Moscow, "Science", 1986.

5. A. Busbom, H. Elders-Boll and H.D. Schotten. Uniformly Redundant Arrays. Experimental Astronomy 8: 97-123, 1998.

6. G. Hermen. Projection image restoration. Fundamentals of reconstructive tomography, translation from English by L.V. Barabin and A. B. Meshcheryakov edited by L. M. Soroko, Moscow, Mir, 1983.

Claims (19)

1. A method of obtaining an image of a given region of an object, comprising scanning in time a given region of an object by irradiating an object with a beam of penetrating radiation and measuring in time the intensity of penetrating radiation after interacting with the object to obtain a signal corresponding to this intensity and then processing it, characterized in that said beam irradiate the entire specified area of the object, and scanning over time of a given area of the object is carried out by changing the distribution time and the radiation intensity in the section of the irradiating beam in such a way that the received signal in time can be described as a convolution and / or Radon transform over the image of a given area of the object, and processing the received signal involves performing reverse convolution and / or inverse Radon transform on it. 2. The method according to claim 1, characterized in that said convolution is a cyclic convolution, and said inverse convolution is a reverse cyclic convolution. 3. The method according to claim 2, characterized in that the cyclic convolution corresponds to convolution with a function having one maximum and flat side lobes in its discrete autocorrelation function. 4. The method according to any one of claims 1 to 3, characterized in that the penetrating radiation is X-ray radiation. 5. The method according to any one of claims 1 to 3, characterized in that the measurement of the intensity of the penetrating radiation after interacting with the object is carried out discretely in time. 6. The method according to claim 5, characterized in that the penetrating radiation is x-ray radiation. 7. A device for acquiring an image of a given area of an object, containing a source of penetrating radiation and at least one detector for measuring the intensity of penetrating radiation after interacting with the object and generating a signal corresponding to this intensity, characterized in that the source of penetrating radiation is configured to form a beam for simultaneous irradiation of the entire specified area of the object, in addition, the device contains a scanner for changing the intensity distribution over time radiant radiation in the cross section of the specified beam so that the signal generated by at least one detector corresponding to the intensity of the penetrating radiation after interacting with the object in time can be described as a convolution and / or Radon transform over the image of a given region of the object, the device also contains a processor for performing the operation of reverse convolution and / or reverse Radon transform on the signal generated by at least one detector corresponding to the intensity radiation after interaction with the object. 8. The device according to claim 7, characterized in that the scanner comprises a screen inhomogeneous for the penetrating radiation and means for moving this screen in at least one direction. 9. The device according to claim 8, characterized in that the screen inhomogeneous for penetrating radiation is a drum having a one-dimensional pattern of length M × N, repeated along the height of the drum N times, with a shift by length M for each subsequent repeat in height, and means for the movement of the specified screen in at least one direction is made with the possibility of rotation of this drum around its axis. 10. The device according to claim 8, characterized in that the means for moving the screen included in the composition of the scanner is inhomogeneous for the penetrating radiation made with the possibility of movement of this screen in two directions. 11. The device according to claim 10, characterized in that the screen, inhomogeneous for penetrating radiation, is a drum with a two-dimensional pattern deposited on its surface, cyclically repeating in height and circumference of the drum, and means for moving said screen in two directions is made to rotate this drum around its axis and moving along it. 12. The device according to claim 10, characterized in that the screen inhomogeneous for penetrating radiation represents a sphere having a two-dimensional pattern on its surface, applied cyclically relative to the orthogonal axes passing through its center, and the means for moving the specified screen in two directions is made with the possibility rotation of this sphere around the indicated axes. 13. The device according to claim 10, characterized in that the screen, heterogeneous for penetrating radiation, is a drum with a one-dimensional pattern applied cyclically around its circumference, and the means for moving said screen in two directions is configured to rotate this drum simultaneously around its axis and around its axis orthogonal to her. 14. The device according to claim 10, characterized in that the screen non-uniform for the penetrating radiation is a disk, on the surface of which a one-dimensional pattern is applied relative to the center of the disk, and the means for moving the said screen in two directions is made with the possibility of rotation of this disk around its center and rotation the center of the disk in a circle lying in the plane of the disk. 15. The device according to claim 7, characterized in that the scanner contains two screens that are heterogeneous for the penetrating radiation and means for moving each of the screens in its direction. 16. The device according to p. 15, characterized in that one of the two screens inhomogeneous for the penetrating radiation is a drum with a one-dimensional pattern cyclically applied around its circumference. 17. The device according to p. 15, characterized in that the screens inhomogeneous for penetrating radiation are two drums placed one in another with orthogonally oriented axes and one-dimensional patterns cyclically applied around their circumferences, and the means for moving said screens is arranged to rotate these drums around specified axes. 18. The device according to p. 15, characterized in that the screens inhomogeneous for penetrating radiation are two drums coaxially placed in each other with one-dimensional patterns cyclically applied around their circumference, the patterns of these drums being inclined at opposite angles of 45 ° relative to their common axis and the means for moving said screens is configured to rotate each of these drums around their common axis. 19. The device according to any one of paragraphs.7-18, characterized in that the source of penetrating radiation is a source of x-ray radiation.

Images