ОПТИКА И СПЕКТРОСКОПИЯ, 2009, том 106, № 6, с. 1014-1024
ЛАЗЕРЫ И ИХ ПРИМЕНЕНИЕ
УДК 621.373:375
LOW-COST ACTIVE OPTICAL SYSTEM FOR FIRE SURVEILLANCE
© 2009 r. A. B. Utkin*, A. V. Lavrov*, and R. M. Vilar**
*INOV - Inesc Inovag&o, 1000-029, Libson, Portugal **DEMAT, Instituto Superior Técnico, Universidade Técnica de Lisboa, 1049-001, Lisbon, Portugal
E-mail: andrei.utkin@inov.pt Received December 14, 2008
Detection of smoke plumes using active optical sensors probides many advantages with respect to passive methods of fire surveillance. However, the price of these sensors is often too high as compared to passive fire detection instruments, such as infrared and video cameras. This article describes robust and cost effective diode-laser optical sensor for automatic fire surveillance in industrial environment. Physical aspects of the sensing process allowing to simplify the hardware and software design, eventually leading to significant reduction of manufacturing and maintenance costs, are discussed.
PACS: 42.62.-b, 42.68.Wt
1. INTRODUCTION
In the area of fire surveillance, active optical sensing technology, based on the backscattered radiation measurement technique similar to LIDAR (LIght Detection And Ranging), presents significant advantages in comparison with the passive methods, such as infrared and video imaging [1-3] and spectroscopic methods [4], in particular, low dependency on light and weather conditions. Detecting smoke rather than flames, active optical sensors do not require line-of-sight to the fire origin, thus allowing for surveillance in complex environments, such as hilly terrains and crammed storage yards. Good directionality and high-precision distance measurements enable lidar to locate accurately the smoke. The detection range is restricted only by the laser-pulse energy and—for distances exceeding ~10 km—by the beam jitter resulted from atmospheric turbulence.
For automatic fire detection, the pattern in the back-scattered-intensity signal, corresponding to the smoke plume must be promptly recognized by an adequate procedure against a background containing spurious peaks due to noise and other targets. Here the lack of knowledge on the local randomly distributed parameters, such as the atmospheric scatter density and the refraction-index inhomogeneities, indicates harnessing artificial-intelligence techniques rather than parametric models. The artificial intelligence methods have been successfully applied to automatic fire detection based on images obtained using ground-based, airborne, or spaceborne infrared or video cameras [1, 5-7]. They also found a wide use in classification of radar [8], sonar [9], and sodar [10] signals. However, these developments are distinct from the application in question because in the image analysis the two-dimensional scenes are classified while in the fire detection the classification deals with a one-dimensional signal: the distribu-
tion of the backscattered radiation intensity along the laser beam propagation direction. The work of Bhatta-charya et al. [11] is an example of application of artificial intelligence algorithms (more specifically, neural networks) to the classification of lidar signals.
Monitoring of smoke emitted by power plants and factories was among the first lidar applications [12, 13], and since those early experiments the interest in lidar has steadily increased. Coupled with sophisticated algorithms for signal processing [14-17], the lidar methods are now widely used for atmospheric research and surveillance [18]. The reported investigations related to smoke detection are mostly concerned with large phenomena on both spatial and temporal scales, such as on-ground and airborne probing of smoke clouds resulting from large forest fires [19] and weapon firing exercises [20], tracking of oil smoke plumes [21], measuring fire smoke density in the stratosphere [22], investigating the correlation between smoke and ozone concentration [23], and characterizing the impact of biomass burning activities on the tropospheric aerosol [24]. Low-height observations are predominantly focused on plumes emitted by power plants [25, 26]. Although smoke detection by lidar has been widely investigated, considerable effort is still required to create effective, reliable, and low-cost methods for fire surveillance.
The influence of lidar parameters on the efficiency of the lidar-assisted fire detection was first investigated in depth, both theoretically and experimentally, by An-dreucci and Arbolino [27]. Later Vilar and Lavrov [28] developed a numerical model for lidar fire detection and used this model to assess the influence of lidar parameters on the monitoring efficiency. Earlier experiments carried out by the authors confirmed that small fires with a burning rate of about 0.02 kg of wood per second can be promptly detected from a distance of 6.5 km [29] (no tests were performed in greater distanc-
Transmitter
Beam forming optics
Laser
Alarm signal and target coordinates
Fig. 1. Schematics of the direct-detection lidar.
es due to difficulties in arranging longer protected horizontal atmospheric path).
The main limitation to the wide use of active optical methods in fire surveillance systems is the high cost of existing sensors. This article reports the development of a low-priced optical system for active fire surveillance using the direct-detection method based on a 905-nm diode laser and a receiver constructed around an avalanche-photodiode (APD) detector. Physical principles of the detection process are analyzed in order to establish the trade-offs between efficiency and cost.
2. BASICS OF FIRE DETECTION USING LIDAR TECHNOLOGY
A direct-detection lidar instrument (Fig. 1) consists of a radiation transmitter (a pulsed laser and beam formation optics) and a radiation receiver (light gathering optical train, photodetector and, in some cases, preamplifier). The transmitter produces short and intense radiation pulses that propagate through the atmosphere, hitting eventual targets (aerosol, smoke, solid obstacles, etc.) and being scattered in all directions; a part of the radiation is backscattered and collected by the receiver, where its power is converted into an analog electric signal, which is amplified, digitized by a data-acquisition unit and recorded as a function of time. Automatic surveillance is provided by a fire-signature recognition system, performing classification of the recorded signals and issuing, if needed, an alarm signal containing information about position of the target that caused the alarm situation.
The distance between the sensor and the target R may be calculated from the time delay t between the laser-pulse emission and the reception of the backscattered signal, R = ct/2, were c is the velocity of light. The raw backscattered-intensity signal S is the receiver-unit output voltage recorded during some period of time immediately after the moment of laser-pulse emission (t = 0). As far as the transition t —► R is merely a re-scaling, the raw lidar signal is conventionally repre-
sented as a plot of S versus the distance R rather than the time t:
S(tS(R) = G/ubph(R) + So, R = ct/2, (1)
where G is the total electronic gain, /ubph(R) = ^phPr(R) is the unbiased photodetector current, qph the photode-tector responsivity, Pr the retroreflected radiation power, and S0 the background component, accumulating all types of electric displacement and low-frequency noise, which can be assumed to be constant during the relatively short measurement time: about 67 ^s for a range of 10 km, according to relation (1).
The retroreflected power Pr can be estimated theoretically using the lidar equation
- ^ ß( R)A r
2 R
Pr(R ) = Ei-^-^ —2 TtrTrecexp
-2 Ja( R ' ) dR'
V o J
(2)
where El is the output laser pulse energy, P the back-scattering coefficient of the medium, Arec the effective receiver area, Ttr and Trec the transmitter and receiver efficiencies, and a the extinction coefficient.
At an early stage of a fire, the characteristic spread of the smoke plume in the laser-beam propagation di rection ARsp (Fig. 2) is about 8-11 m. As follows from previous works [30-33], to reveal the specific structures that make the smoke-plume signatures different from other signal peaks, the data-acquisition unit must digitize the photodetector output with a sampling interval 8R ~ 0.75 m, providing at least 10-15 signal points within the smoke-plume peak. Eventually, we deal with the discrete-time lidar signal in the form
S( R ( t)) = Co exp R
i
R
-2ja( R ) dR'
V 0
+ So,
C0 — G ^ph El 2 Arec Ttr Trec = c°nSt,
Fig. 2. Spatial parameters critical for smoke-plume detection.
R, m
Fig. 3. Composition of the raw lidar signal.
digitized at the points
ti = 2Ri/c, Ri = i5R, i = 0, 1, ..., i'max,
imax = Rmax/5R .
(4)
The smoke-plume retroreflection peaks are detected against a background contaminated by electronic and atmospheric noise (Fig. 3). Usually electronic noise does not depend on the distance and can be estimated from a signal segment recorded far beyond the range of the instrument, where no retroreflection signal is expected [34]. Apart from this unstructured noise, the li-dar signal may contain peaks due to retroreflection from dust clouds, hills, trees, buildings, etc. Since backscattering occurs in a thin surface layer, the solidtarget signatures are represented by narrow pulse-like waveforms, whose shape is mainly defined by the
bandwidth of the detection channel and the rate of the analog-to-digital conversion.
3. METHODOLOGY OF AUTOMATION 3.1. Characterization
The lidar technology has a potential to locate targets with the precision of a few meters. The angular target position (the azimuth 9 and elevation see Fig. 2) is given by the laser beam direction, but the distance to the smoke plume Rsp must be the calculated from the backscatter signal.
According to Eq. (3), the shape of the smoke-plume signatures in
Для дальнейшего прочтения статьи необходимо приобрести полный текст. Статьи высылаются в формате PDF на указанную при оплате почту. Время доставки составляет менее 10 минут. Стоимость одной статьи — 150 рублей.