DEVICE FOR READING INFORMATION FROM X-RAY FILM

V. P. Abramov, V. I. Gorbunov, B. N. Epifantsev

INTRODUCTION

Increasing attention has been, and continues to be, devoted to the automation of the process of quality control of materials and finished products by means of penetrating radiation. These efforts began to be carried out especially intensively with the introduction into industry of introscopes and flaw detectors based on the scintillation method. However, the main direction of these efforts comes down to automating either the process of recording changes in radiation intensity or the various possible movements of the flaw detector [7]. At the same time, such an important and, perhaps, fundamental section of any inspection method as the analysis and decoding of flaw-detection information has up to the present time remained the monopoly of the human operator. Meanwhile, the present level of development of science—in particular, the section of technical cybernetics concerned with pattern recognition—makes it possible to pose and solve such problems associated with the analysis and classification of incoming information.

The problem of automatic analysis and recognition of images of defects on radiographic pictures, in order to be solved, requires the development of many issues, one of which is the choice of the principle for converting visual information into electrical signals, i.e., in essence, the development of the sensing device of a recognizing automaton. As is known, the sensing device of a recognition system performs the function of measuring feature values, and, consequently, the requirements imposed on it are limited by the necessary accuracy, resolution, and speed of operation. With the aid of a scanning system, such features as the number and coordinates of objects, their configuration, color, mutual arrangement, linear dimensions, etc., can be “extracted” from the observed image. Thus, at the output, electrical signals are obtained that contain information about the various features and properties of the image being analyzed.

In the present article, questions related to the choice of the principle for converting visual information into electrical signals for a specific application are considered, and the basic relations for the lighting-engineering calculation of a scanning device are given in order to ensure a specified signal-to-noise ratio.

Principles of Scanning

In terms of the operations performed, all scanning systems may be subdivided into:

1 — those performing passive field scanning, carried out according to a preset program;

2 — those performing active field scanning, the program of which changes during the scanning process depending on the particular field pattern.

Both active and passive systems can survey the field either continuously or discretely, with interruptions.

A region of a plane field can be completely covered by arbitrary scanning trajectories, but in engineering the most widespread are line-by-line and circular scans, since they can be implemented with fairly simple devices.

The type of scanning, as well as the scanning speed of the object, is determined by the task at hand and by the dynamic properties of the object.

An analysis of currently developed scanning systems makes it possible to identify the following principles of their construction [1, 2].

1. The use, for scanning purposes, of ordinary optical devices aimed at a single point of the object and mounted on a special mechanical device. In this case the latter changes the spatial position of the optical axis of the system in such a way that the object is scanned along the required trajectory. The principal drawback of such devices is the low permissible scanning speed, since the spatial motion is performed by the entire device, which has a comparatively large mass.

2. Scanning systems in which image scanning is carried out by means of an optomechanical device. Thus, works [3, 4] describe scanning devices with an optomechanical generator of the light spot. They provide high resolution and luminous intensity; however, because of the complexity of the mechanical system, bulkiness, and difficulty of controlling the position of the light spot, such systems have found little use. The well-known Nipkow disk also belongs to this type of system.

3. Photoelectronic scanning systems, in which image scanning is carried out, as a rule, by deflecting a stream of photoelectrons with external magnetic fields. For image analysis, a number of widely used transmitting television tubes have been developed (iconoscopes, supericonoscopes, orthicons, vidicons, dissectors, etc.). The positive properties of photoelectric vacuum scanning devices are their high sensitivity, resolution, and dynamic characteristics. However, some of their shortcomings are due to their great complexity, the presence of units that form signals of a special shape supplied to the deflection systems, and their high sensitivity to vibration and shock loads.

A separate class is made up of devices in which scanning is carried out by moving special screens in front of the sensitive element; through the apertures of these screens, radiation from different areas of the image is fed in succession. However, these systems have low sensitivity.

Recently, non-vacuum photoelectronic scanning devices have also been developed, based on the use of the phenomenon of electroluminescence [5]. In them, scanning is carried out by alternately applying, along mutually perpendicular ...

transparent electrodes of electrical signals that unlock different sections of the electroluminophore screen. The scanning element of such devices is sufficiently simple; however, its sensitivity and resolving power are low. Scanning devices with fiber-optic elements, which recently have become increasingly widespread, may also be assigned to this same group. Their use makes it possible to create complex scanning systems on the basis of simple switching and scanning photoelectronic devices with circular scanning trajectories.

Photoelectronic and electroluminophore scanning devices possess almost equivalent positive properties. They have approximately the same sensitivity and the same dynamic properties. However, the resolving power of photoelectronic scanning devices is considerably higher than that of electroluminophore devices. The principal advantage of electroluminescent devices is the simplicity of the scanning element and its resistance to vibration and shock loads.

A comparative analysis of the considered principles for constructing scanning systems, as well as the stated requirement of developing a high-speed recognizing automaton that deciphers 500,000 images per year with high sensitivity and resolving power, predetermines the choice of a photoelectronic scanning device.

To select an image-viewing mode, let us consider the possible variants presented in Fig. 1 (a, b, c, d).

The specificity of the problem posed—deciphering radiographic images—requires the solution of two successive tasks: 1) detection of defects recorded on the X-ray film, 2) recognition or classification of the detected defects. In this connection, the most expedient principle to be adopted in developing the scanning device is sequential viewing of a specified field. Thus, analysis of the principles of construction of scanning devices and the requirements that follow from solving the problem of deciphering radiographic images make it possible to choose a photoelectronic scanning device operating in the mode of sequential viewing of a specified field. The widest application among such devices has been obtained by systems with a running beam, in which a light spot on the screen of an ELT is used as the light source, and a PEM serves as the receiver. They provide high resolving power and high speed, which

1. Fig. 1a. Sequential viewing of all parts of the image by the field of view.
   Fig. 1b. Movement of the field of view along the boundary or along the contour of the image (tracking scan).
   Fig. 1c. Viewing the most informative parts of the image.
   Fig. 1d. Transition from scanning the entire image to movement of the field of view along the contour of the image (mixed mode).

is determined by the afterglow time of the tube screen, and they make it possible to correct the output signal over the field of view comparatively simply.

Signal-to-noise ratio and basic relationships for the photometric design of the scanning device

The schematic of a device for reading information recorded on X-ray film is shown in Fig. 2.

The light source is a sharply focused spot on the screen of scanning tube 1. By means of deflecting devices this spot moves over the surface of the screen, describing a raster of the required format. The image of the luminous spot is projected by objective 2 onto X-ray film 3.

1. Fig. 2. Schematic of the reading device.

The light flux that has passed through the X-ray film is collected by condenser 4 on the photocathode of electron multiplier 5. The magnitude of the current at the output of the photomultiplier at each instant is proportional to the transparency of the film at the point illuminated by the spot. Since the luminous spot successively traverses different parts of the surface of the X-ray-film frame, the output of the photomultiplier yields a time-varying video signal from the entire scanning field, which is amplified and fed to the information-processing unit.

Since the efficiency and reliability of the scanning device as a whole depend on the correct choice of the parameters of its basic elements—the CRT and the photomultiplier—the most important of these are briefly noted below; more detail on this can be found in [8, 9].

Requirements for the CRT. First, a short afterglow time of the screen is required (from fractions to units of microseconds). With a long afterglow time, the sharpness and contrast of the image in fine details in the line direction will be appreciably reduced. Second, a high resolving power, determined by the dimensions of the working part of the screen and of the light spot, is required. Third, a high brightness of the spot in the tube and a good signal-to-noise ratio in the video signal are required.

As for the “light-signal” converters—the photomultipliers—they are required to have a high integral sensitivity of the photocathode, referred to the spectral density of the radiation of the screen phosphor:

\[ \delta^{*} = \delta_{\max}\cdot \kappa_{m}, \tag{1} \]

where \(\delta_{\max}\) is the maximum absolute spectral sensitivity of the photocathode.

\[ \kappa_{m} = \frac{\displaystyle \int_{0}^{\infty} \delta_{\lambda} p_{\lambda}\, d\lambda} {\displaystyle \int_{0}^{\infty} P_{\lambda}\, d\lambda} \tag{2} \]

— the reduction coefficient for the relative spectral characteristics of the photodetector \(\delta_\lambda\) and the radiation source \(P_\lambda\). The values of \(\kappa_m\) for various combinations of phosphor and photocathode types have been tabulated [9, 12].

In turn, the absolute spectral sensitivity is determined by

\[ \delta_{\max}=\delta_{\mathrm{int}}\cdot v_{\max}\cdot \kappa_2, \tag{3} \]

where \(\delta_{\mathrm{int}}\) is the integral sensitivity given for PMT photocathodes in handbooks.

\[ \kappa_2= \frac{\displaystyle\int_0^\infty v_\lambda\cdot r_{\mathrm{el}}\,d_\lambda} {\displaystyle\int_0^\infty \delta_\lambda r_{\mathrm{el}}\cdot d_\lambda} \tag{4} \]

— the conversion coefficient. Its values for photocathodes of various types have been tabulated [9, 12]. \(V_{\max}\) is the maximum spectral sensitivity of the eye, corresponding to \(\lambda=555\) nm.

The synthesis of the scanning device should be carried out on the basis of ensuring the required signal-to-noise ratio. The output signal taken from the load of such a device is subject to substantial fluctuations. Experimental study by a number of authors of the sources of the resulting noise and of its character has shown that the most important of these are the noise of the photomultiplier itself and variations in the intensity of screen luminescence. The latter can be considerably reduced by supplying the tube from stabilized voltage sources, and also by applying a local narrow-band feedback circuit from the PMT output through an amplifier and a filter to the tube grid, which controls the brightness; as a result, the intensity of the spot luminescence can be stabilized within \(\pm 1\%\) [10].

As is known, the signal-to-noise ratio is understood as the ratio of the absolute value of the excess of the mean signal level \(U_c\) over the mean background level \(U_\phi\) to the rms value of the noise.

\[ \psi=\frac{U_\phi-U_c}{\delta_{\mathrm{sh}\phi}} =\frac{K U'_\phi}{\delta_{\mathrm{sh}\phi}}, \tag{5} \]

where

\[ K=\frac{U_\phi-U_c}{U'_\phi} \]

is the contrast coefficient of the information carrier, characterizing the brightness distribution on the carrier, and \(\delta_{\mathrm{sh}\phi}\) is calculated, taking into account the dark current of the PMT, the components of thermal noise, and the amplifier noise, by the formula [9]:

\[ \delta_{\mathrm{sh}\phi}= \sqrt{ 2e\,(I+I_t)\,\Delta f\,f(G)\,R_\phi^2 + \frac{4kT\,\Delta f\,(R_\phi+R_{\mathrm{eqv}})} {|G(\omega)|^2} }. \tag{6} \]

Formula (5) can be rewritten, using (6) and taking into account that the response of the PMT can be represented as \(U_c=I_cG\cdot R_\phi=\delta^*\Phi_kG\cdot R_\phi\), in the following form (neglecting the thermal noise components):

\[ \psi= \frac{\kappa \delta^*\Phi_k} {\sqrt{2e\Delta f f(G)\left(I_t+\delta^*\Phi_k\right)}}. \tag{7} \]

If \(I_t\ll \delta^*\Phi_k\), then

\[ \psi=\kappa\sqrt{\frac{\delta^*\Phi_k}{2e\Delta f f(G)}}. \tag{8} \]

where \(\Phi_k\) is the luminous flux incident on the PMT photocathode;

\(\Delta f\) — passband of the device;
\(f(G)\) — noise excess factor;
\(e\) — electron charge.

One of the most important characteristics of a photoelectronic scanning device is the magnitude of the minimum permissible flux at which the specified signal-to-noise ratio is realized. Using formula (8), we obtain:

\[ \Phi_{\mathrm{k}}=\frac{\psi^{2}\cdot 2e\Delta f f(G)}{\kappa^{2}\delta^{*}} . \tag{9} \]

In turn, the luminous flux \(\Phi_{\mathrm{k}}\) is related to the total flux on the CRT screen \(\Phi_{\mathrm{p}}\) through the transmission coefficient of the radiant flux.

\[ \Phi_{\mathrm{k}}=\kappa_{\mathrm{d}}\Phi_{\mathrm{p}}=\kappa_{\mathrm{d}}\gamma I_{\mathrm{l}}U_{\mathrm{a}}, \tag{10} \]

where

\(\gamma\) — efficiency of the phosphor;
\(I_{\mathrm{l}}, U_{\mathrm{a}}\) — current and anode voltage of the CRT.

For the scheme shown in Fig. 1, the coefficient \(\kappa_{\mathrm{d}}\) may be represented as

\[ \kappa_{\mathrm{d}}=\kappa_{\mathrm{d}_1}\cdot\kappa_{\mathrm{d}_2}, \]

where \(K_{\mathrm{d}_1}\) is the coefficient of the optical system, characterizing the part of the luminous flux that has passed through the objective and falls on the carrier; it is determined as:

\[ \kappa_{\mathrm{d}_1}= \frac{\kappa_{0}} {1+4(x^{2})\left(\dfrac{M+1}{M}\right)^{2}}, \tag{11} \]

where \(\kappa_{0}\) is the transmission coefficient of the objective;
\(x=F/D\) is the coefficient inverse to the relative aperture of the objective;
\(M\) is the linear magnification of the system (the ratio \(a'/a\));
\(\kappa_{\mathrm{d}_2}\) is the coefficient characterizing the part of the luminous flux that has passed through the carrier and falls on the photocathode, equal to

\[ \kappa_{\mathrm{d}_2}=\rho\kappa_{\mathrm{k}}, \tag{12} \]

where \(\kappa_{\mathrm{k}}\) is the transmission coefficient of the condenser;
\(\rho=0.1^{D}\) is the transmission coefficient of the carrier;
\(D\) is the optical density of the film.

Analyzing formulas (10) and (12), and taking into account the fact that the magnitude of the luminous flux of the spot on the CRT screen is limited by the accelerating voltage, the type of phosphor, its efficiency, and the maximum permissible load, it may be stated that provision of the required operating mode of the PED in terms of average current (and hence of a reliable signal-to-noise ratio) should be achieved through the correct choice of the transmission coefficient of the radiant flux \(\kappa_{\mathrm{d}}\), which in turn is attained by a rational design of the reading device.

An equally important characteristic for such devices is the effect of afterglow, which imposes limitations on the speed of response of the device, since the magnitude of the additional flux \(\Delta\phi\) is proportional to the scanning speed \(V\). Therefore, in calculating and designing scanning devices, this relationship must be kept in mind, since an excessive increase in the scanning speed may disrupt the normal operating mode of the device. A detailed consideration of this question may be found in [11].

In conclusion, we shall present some technical characteristics of the scanning device that has been developed.

As the light source, an ELT tube of type 18L047A was used, as the tube having the shortest afterglow time among oscillographic tubes. The “light-signal” converter is an FEU-31 photomultiplier. Brightness control is carried out by means of local narrow-band feedback from the FEU output.

In the optical system projecting the beam onto the X-ray film, a “Jupiter-3” lens is used, with a resolving power of 30 lines/mm at the center and 14 lines/mm over the field, and with good speed (relative aperture 1:1.5). The exit pupil of the lens is projected onto the FEU photocathode by a condenser with focal length \(f = 30\) mm and relative aperture 1:1.

The output signal of the scanning device is encoded by a 10-bit voltage-to-code converter and is sent for subsequent processing to the ETsVMM-20.

References

1. G. P. Katys. Scanning Photoelectric Search and Tracking Devices, Nauka Publishing House, Moscow, 1964.

2. G. P. Katys. Automatic Survey and Search in the Optical Range, Nauka Publishing House, Moscow, 1966.

3. P. M. Blackal, G. M. Macleod, P. Zannels. The automatic analysick of spark chambers using a flying spot digitister. Meeting on Programming for HRD and other flying spot devices held at the Callege de France, Paris, 1963.

4. W. F. Baker, A. Spark chamber film reading sqstem. Meeting on Programming for NRD and other flying spot devices held at the College de France Paris, 1963.

5. I. R. Lyashchenko. Method and Device for Transmitting a Television Image, Author’s Certificate No. 115236, 1955.

6. Reading Devices, collection of articles, VINITI, Moscow, 1965.

7. V. I. Gorbunov, N. Ya. Makarov, V. V. Chemev, V. P. Abramov, L. B. Voroshen. Automatic quality control of large-thickness products. Proceedings of the 5th Conference “Automatic Control and Methods of Electrical Measurements,” vol. II, Nauka Publishing House, Siberian Branch of the Academy of Sciences of the USSR, Novosibirsk, 1965.

8. A. I. Petrenko. Conversion of Graphs into Electrical Signals, Naukova Dumka Publishing House, Kiev, 1964.

9. A. I. Petrenko. Automatic Input of Graphs into Computers, Energiya Publishing House, Moscow, 1968.

10. Reading Devices, VINITI collection, Moscow, 1962.

11. I. Ya. Butlitskii. Lighting-engineering calculation of a television-signal sensor with a running beam, Tekhnika kino i televideniya, No. 5, 1957.

12. V. K. Eliseev. Lighting-engineering calculation of photoelectric image converters. Proceedings of the seminar “Pattern Recognition and the Design of Reading Automata,” issue 1, Kiev, 1966.

13. R. I. Rybak, O. F. Purin. Photoelectric converter of a device for processing optical information by means of a computer. Proceedings of the seminar “Pattern Recognition and the Design of Reading Automata,” issue 2, Kiev, 1968.