Vibrating fiber-optic detection devices.
Vibrating fiber-optic detection devices are designed to detect an intruder by the vibrations of the alarm barrier they create during penetration into the protected facility.
The operating principle of vibrating fiber-optic detection devices is based on recording mechanical vibrations or movements of the barrier that occur when an intruder attempts to destroy or overcome the physical barrier.
The sensitive element of such systems is a specially developed fiber-optic cable that converts local deformations of the cable that occur during mechanical vibrations into changes in the characteristics of the laser radiation passing through the optical fiber. The cable is attached either directly to the barrier or to a special light metal canopy above it. Changes in the characteristics of the laser radiation are recorded by a signal processing unit, which, in accordance with a specified algorithm, issues an alarm signal. In addition to the signal processing unit, the vibration fiber-optic detector includes an optical quantum generator and a monitor.
The functional diagram of the detector is shown in Fig. 1. With regard to the vibration fiber-optic detector, its elements are:
The external impact is mechanical vibrations or movement of the barrier that occur when an intruder attempts to destroy or overcome the physical barrier.
The sensitive element of the detector is the fiber-optic cable.
The signal processing unit is present.
The output device is usually implemented as an output relay.
In addition, the detector additionally includes:
an optical quantum generator,
a monitor
Therefore, the functional diagram of the vibration fiber-optic detector will look like (Fig. 2.):
Device
Sensitive element
Fiber optic cables, which are usually used for transmitting information, can also be used as sensors for perimeter security systems. Deformation of the optical fiber changes its optical parameters and, as a consequence, the characteristics of the radiation passing through the fiber. Due to the specificity of the physical principles used, fiber optic systems are characterized by very low susceptibility to electromagnetic interference, which allows them to be used in unfavorable electrophysical conditions.
Optical fiber is generally a coaxial light guide. Light travels along the central part (core) of the cable. Adjacent to the fiber core is a transparent cladding, which has a lower refractive index than the core. Light traveling at an angle to the fiber axis is reflected from the interface between the core and the cladding and is concentrated in the central part of the fiber. The outer opaque coating serves to mechanically protect the cable.
Fiber optic cables are divided into multimode and single-mode.
Mode – a type of path along which light can propagate.
The number of modes allowed by a fiber optic cable typically ranges from 1 to 100,000. Thus, the fiber allows light to propagate along many paths, the number of which depends on the size and properties of the fiber.
Core diameter of multimode fibersis usually 50-100 microns. A large number of wave types (modes) with different geometric parameters propagate simultaneously through such a fiber. These rays experience multiple reflections from the boundary between the core and the cladding, which leads to noticeable attenuation of the signals (Fig. 2.29). Core diameter of single-mode optical fibersis no more than 10 microns. Only one type of wave (mode) can propagate in such a light guide, and the attenuation of light here is significantly less than in multimode light guides (Fig. 3).
Miniature semiconductor lasers or LEDs are usually used as a radiation source.
At the cable output, the radiation is recorded by a photodetector, which converts the optical signal into an electrical signal. When the fiber is deformed, the internal reflection conditions change, which results in changes in the phase and spatial characteristics of the beam at the cable output. These changes are recorded by the photodetector and processed by the signal analyzer.
The refractive index of the optical cladding is less than 1% less than the refractive index of the core. The characteristic values of the refractive indices are N = 1.47 for the core and N = 1.46 for the optical cladding. Fiber manufacturers strictly control the difference in indices to obtain the desired fiber characteristics.
Methods of recording penetration signals
Method of recording intermode interference
A semiconductor laser usually generates several dozen modes (spectral lines) close in frequency with a certain energy distribution over the emission spectrum. If a multimode fiber optic cable is subjected to mechanical effects, then at its output the emission spectrum registered by the receiver changes due to the appearance of losses of radiation energy from microbending of the cable, which allows detecting cable deformations. Losses from microbending are losses of optical energy caused by the escape of light beyond the fiber due to local changes in the profile of the core/optical cladding boundary. Fig. 4 shows that these changes in the core/optical cladding boundaries can lead to reflection of high-order modes at angles that do not allow further reflections. In this case, the light leaves the fiber.
Snell's law establishes the critical angle of incidence from the ratio between the refractive indices of the two media in which the light propagates.
At angles of incidence greater than the critical angle, the light is completely reflected from the interface between the two media and remains in the optical fiber. But if, due to cable deformation, the angle of incidence of the light beam becomes less than the critical angle, then part of the beam will overcome the interface between the two media and exit the core of the optical fiber.
Speckle structure registration method
At the output of a multimode optical fiber, a so-called speckle structure is observed, which is an irregular system of light and dark spots. When the fiber is deformed or vibrated, the speckle structure of the radiation undergoes changes. Spatially sensitive photodetectors are used here to detect cable deformations.
The reception efficiency, which can be defined as the ratio of the power of the speckle participating in the formation of the registration signal to the total power of the incident radiation, is determined by the spatial coherence of the radiation. The spatial coherence of the speckle is characterized by the coherence radius: the statistically average radius of the coherence spots — areas on the wave front with a regular phase change. This limits the input angular aperture, which is achieved using an optical receiving device made of a lens. The registration signal can be increased by increasing the power in the coherence spot and by reducing the size of the light spot on the photocell. The graphical interpretation has the form shown in Fig. 5
Interference method
This method uses the principle of two-beam interferometry. The laser beam is split into two and directed into two identical single-mode optical fibers, one of which is the detecting fiber and the other is the reference fiber. At the receiving end, both beams form an interference pattern. Mechanical impacts on the sensitive cable lead to changes in the interference pattern, which are recorded by a selective photodetector (Fig. 6).
Light interference is the addition of several light waves, resulting in alternating light and dark areas, i.e. a redistribution of the energy of these waves in space (along the wave front) occurs.
Two monochromatic light waves, superimposed on each other, excite oscillations of the same direction at a certain point in space:
, ,
where — are the phase velocities of the first and second waves, respectively.
E = E1 + E2
The phase difference of the oscillations excited by the waves at point M is
,
where is the optical path difference.
λo is the wavelength
the width of the frequency spectrum of monochromatic radiation.
Interference conditions:
maximum, oscillations occur in the same phase (where m = 0, 1, 2, …);
minimum, oscillations occur in antiphase (where m=0, 1, 2, …).
An interference pattern is produced only by coherent waves (matched), waves of the same wavelength (or frequency), which arrive at a given point with a constant phase difference over time. Monochromatic waves satisfy this condition – waves of one specific and strictly constant frequency that are unlimited in space. A distinction is made between temporal and spatial coherence. Any non-monochromatic light can be represented as a set of independent harmonic trains that replace each other. A wave train is an intermittent emission of light by atoms in the form of separate short pulses. The average duration of one train – coh is called the coherence time, which cannot exceed the emission time t. During this time, the wave propagates in a vacuum over a distance called the coherence length (train length).
Two sources whose dimensions and mutual arrangement allow (with the required degree of monochromaticity of the light) to observe interference are called spatially coherent.
The coherence radius (or spatial coherence length) is the maximum distance at which interference can occur.
where: r is the coherence radius,
λ is the wavelength of light,
φ is the angular size of the source.
Light interference can be obtained using a Fresnel biprism. A Fresnel biprism is two prisms connected by their bases with identical and very small (on the order of fractions of a degree) refractive angles.
If such a biprism is positioned so that the direction of light is parallel to the edge of the biprism AB, then a light interference pattern can be obtained on the photodetector.
Monochromatic light waves coming from sources S1 and S2 in the region of DAE create an interference pattern. Figure 7 shows a top view, the selected beams 1 and 2 give a picture of the addition of beams at point M. Depending on the path difference, we have either a dark strip (wave damping) or a light strip (wave amplification).
Signal processing unit
To separate the signals generated by the intruder from noise and interference, a signal analyzer based on the neural network principle is used. The use of a neural network ensures high reliability of detection with a low level of false alarms. An enlarged block diagram of the signal processing unit is shown in Fig. 8.
An adaptive filter is located at the processor input, which optimizes the operation of the device.
The signal is fed to the input of the neural network analyzer after passing through the DSP (digital signal processing) processor, the operating principle of which is based on fast Fourier transform algorithms.
The main signal processing occurs in the logical analyzer, built on the basis of neural networks.
A neural network or neurocomputer is a computing system in which the algorithm for solving problems is presented in the form of a network of threshold elements with dynamically reconfigurable coefficients and tuning algorithms that are independent of the dimensionality of the network of threshold elements and their input space.
To describe neural networks, a special circuit design has been developed in which elementary devices — adders, synapses, neurons, etc. — are combined into neural networks designed to solve problems.
The basic element of a neural network is a neuron. A standard formal neuron (Fig. 9) consists of an input adder, a nonlinear converter, and an output branch point. One of the main elements of a neuron is an adaptive input adder (Fig. 10), which calculates the scalar product of the input signal vector.
The input signal, which at each moment of time is characterized by a set of parameter values xn, is fed to the input of the adder through a system of parallel linear connections-synapses. The number of parameters xn and their physical nature can be different, for example, the signal amplitude, derivative, energy, etc. The parameter values are calculated at the stage of preliminary processing of the real signal to be analyzed.
The synapse receives the value of a certain parameter xn at the input and multiplies it by the adjustable coefficient n.
The nonlinear signal converter receives a scalar output signal Y and converts it to (Y). The branch point is used to send one scalar output signal (Y) to several addresses. The most commonly used nonlinear functions of the neuron converter are threshold, three-valued threshold, and sigmoid functions.
Neural Network Architecture
Among the many neural network architectures, one can single out the basic architecture – layered networks (Fig. 11).
Layered networks: neurons are arranged in several layers (Fig. 2.38). Neurons of the first layer receive input signals, transform them and pass them on to neurons of the second layer via branching points. Then the second layer is activated, and so on, up to the k-th layer, which produces output signals. Unless otherwise stated, each output signal of the i-th layer is fed to the input of all neurons of the i+1-th. The number of neurons in each layer can be any and is not connected in any way with the number of neurons in other layers. The standard method of feeding input signals: each neuron of the first layer receives all input signals. Three-layer networks are especially widespread, in which each layer has its own name: the first is the input, the second is hidden, and the third is the output.
Neuron Functioning
The functioning of a neuron in a neural network occurs as follows. At the current moment in time, the neuron receives signals from outside in the form of values of the selected parameters xn . These signals are called input signals. The signal from each input in the synapse is multiplied by the weight coefficient n of this input (the weight coefficients of the inputs can be different) and is added in the adder with other signals, also multiplied by the weight coefficients of the corresponding inputs.
In a neuron, the signal value changes according to the neuron transformation function . After transformation, an output signal is formed in the neuron, which is transmitted to other neurons. A neural network that receives a certain signal at the input is capable of, after passing it through neurons, producing a certain response at the output, which depends on the weight coefficients of all neurons.
In order to achieve the desired result from the network, it must be trained.
The neural network training algorithm (the so-called backpropagation algorithm) consists of comparing the output of the last layer of neurons with the training sample and, based on the difference between the desired and the actual, a conclusion is made about what the connections of the neurons of the last layer with the previous one should be. Then a similar operation is performed with the neurons of the penultimate layer. As a result, a wave of change in the weights of connections runs along the neural network from the output to the input. In general, a neural network has two remarkable properties: the ability to learn on a certain set of examples and to stably recognize (predict) new situations with a high degree of accuracy, and in conditions of strong external interference, such as the appearance of contradictory or incomplete values. System training is reduced to the operation of the weighting coefficient selection algorithm, which operates without the direct participation of the operator.
The article uses materials from the book «Perimeter Protection Systems».
Authors — G. F. Shanaev, A. V. Leus, under the general editorship of S. I. Korchagin.
The book was published by Security Focus (Moscow).
Detailed information secnews.ru/articles/16357.htm