'HVLJQRIFXELFSKDVHGLVWULEXWLRQOHQVHVIRUSDVVLYH
LQIUDUHGPRWLRQVHQVRUV
a
Giuseppe Antonio Cirino a and Luiz Gonçalves Neto*b
EletroPPAr Ind. Eletrônica Ltda., Av. Dr. Labieno da C. Machado 3370, Garça, S. P. – Brazil
b
University of São Paulo, Av. Dr. Carlos botelho 1465, São Carlos, S.P. - Brazil
$%675$&7
A cubic-phase distribution is applied in the design of inexpensive lenses for passive infrared motion sensors. The
resulting lenses have the ability to distinguish the presence of humans from pets by the employment of the so-called
wavefront coding method. The cubic-phase distribution used in the design can also reduce the optical aberrations
present on the system. This aberration control allows a low tolerance in the fabrication of the lenses and in the
alignment errors of the sensor.
Keywords: Passive infrared sensor, cubic phase distribution, pyroelectric detector.
,1752'8&7,21
Passive infrared (PIR) motion sensors has been extensively used in homeland and corporative electronic security
applications 1. Figure 1 shows the photography of a typical low cost passive infrared (PIR) pyroelectric motion sensor.
The principle from which this device works is based on the employment of a pyroelectric detector, which is responsible
for transducing the incident IR radiation to an electrical signal. This signal is amplified and filtered, generating another
signal to an alarm element. To avoid false detections due to temperature variation, the pyroelectric detector presents a
dual active region, with reverse polarization respect to each other.
Figure 1: Typical low-cost passive infrared (PIR) pyroelectric motion sensor.
Associated with the photo detector and the circuitry, there is a multi zonal Fresnel lens, which monitors different spatial
zones and concentrate the IR radiation from a body that moves within the monitoring volume. In this case, the lenses
are designed only to make sure that the incident IR radiation is concentrated on the pyroelectric detector surface,
regardless the formation of a good quality image. When a human corporal mass (here called by target mass) moves
through the volumetric monitoring field of the sensor, different Fresnel lenses focus the IR radiation on the pyroelectric
*
Phone (55)(16)273-9350, E-mail: [email protected]
detector, generating an electrical signal at the pyroelectric detector output. This signal is then amplified and filtered,
making the sensor to trip, as shown in figure 2. The electrical current in the in the pyroelectric detector is proportional
to the time rate of change of the target mass with temperature T:
I = p( t ) A
∆T
∆t
(1)
where p(t) is the pyroelectric coefficient of the detector, A is the pyroelectric detector electrode area and T is the
temperature of the moving target mass.
Figure 3 shows the photography of a segmented Fresnel lens typically employed in IR motion sensor applications. This
lens presents 24 detection zones, with 3 sets of lenses that monitor 3 different distance fields: a near field (typically 1-2
meters from the sensor), a medium distance field (typically 3-6 meters from the sensor), and a far field (typically 8-15
meters from the sensor).
Figure 2: Schematic view of the principle of a PIR sensor. When a human corporal mass (target mass) moves through the monitoring
volume field, an electrical signal at the pyroelectric detector output is generated, enabling the sensor to trip.
The electrical waveform at the pyroelectric detector output depends on the distance between the detector and the target
mass, on the amount of mass present in the target, and on the speed of the target with respect to the detector. The sensor
operates at the 8-14 micrometers wavelength range, because at this range the contrast between the target mass
temperature and the environment temperature is more accentuated.
Although much progress has been made in this field, a typical problem however, remains unsolved, namely the ability
of the sensor to distinguish the presence of humans targets from pets using only an optical processing. Several
companies have the so-called pet immune sensor, which acts in electronic post-processing way, i.e., the signal is
collected by a pyroelectric detector and processed (filtered) afterwards. A common problem that remains unsolved in
the application of PIR motion sensor is their ability to distinguish the presence of humans from pets using only an
optical processing. To solve the pet immunity problem in PIR motion sensors, a new approach that consists of a hybrid
optical pre-processing, also called wavefront coding 2,3,4, can be applied. In this field, authors have successfully reported
the employment of wavefront coding to correct the aberrations present in incoherent image systems 5,6, bringing them to
be invariant to misfocus within a predefined range (the same considerations can be extended to the chromatic
aberration) 7. Because of the aberration invariance, it is possible to use a low-cost and low-precision optics, to
producing a system that images with the performance of high-cost, high precision, or near diffraction-limited system5 .
In the design of a PIR motion sensor lens, the wave coding approach can help to distinguish the presence of pets from
humans. In this paper we verify the application of the cubic phase distribution to improve the immunity of PIR sensors
to false alarms generated by pets.
Figure 3: Photography of a multi-zonal, segmented Fresnel lens typically used in PIR sensors. There exist 24 detection zones, with 3
sets of lenses that monitor 3 different distance fields: a near field (bottom of the lens), a medium distance field (mid height of the
lens), and a far field (top of the lens). Each segmented lens is delimited by rectangles.
7+(25<2)$%(55$7,21,19$5,$1&(
An incoherent imaging system is linear in intensity and the impulse response, the point spread function (PSF), is the
squared magnitude of the amplitude impulse response h(xi,yi), the amplitude point spread function 8. The variables xi
and yi are the coordinates in the image plane. The optical transfer function (OTF) +(u,v) of the system can be defined
from h(xi,yi) by
+∞
∫ ∫ h(x , y )
i
H( u , v) =
2
i
exp[− j2π(x i u + y i v )]dxdy
−∞
(2)
+∞
∫ ∫ h (x , y )
i
i
2
dxdy
−∞
When an imaging system is diffraction limited, the amplitude point spread function is obtained from the Fraunhofer
diffraction pattern of the exit pupil P(xp,yp), with P(xp,yp)=1 in the region of the exit pupil and P(xp,yp)=0 outside. The
variables xp and yp are the coordinates in the pupil plane. For aberrated imaging systems, the aberrations can be directly
included in the plane of the exit pupil 8. The phase errors in the plane (xp,yp) of the exit pupil is represented by
kW(xp,yp), where k=2π/λ and W is effective path-length error. The complex amplitude transmittance 3(xp,yp) in the exit
pupil is given by:
[
P ( x p , y p ) = P( x p , y p ) exp jkW ( x p , y p )
]
(3)
The complex function 3(xp,yp) may be referred as the generalized pupil function. The PSF of an aberrated incoherent
system is simply the squared magnitude of the Fraunhofer diffraction pattern of an aperture with amplitude
transmittance 3(xp,yp).
The most common approach to obtain a focus-invariant incoherent optical imaging system is to stop down the pupil
aperture or to employ an optical power absorbing apodizer, with possible π phase variations. The aperture can be
viewed as an absorptive mask in the pupil plane of an optical imaging system. Although these methods increase the
amount of depth of field, there is a loss in optical power at the image plane and the reduction of the diffraction-limited
image resolution 2,5,8.
A non-absorptive separable cubic phase mask can be included in the exit pupil to produce an optical point spread
function (PSF) that is highly invariant to misfocus 2,5. Considering the cubic phase distribution, the PSF is not directly
comparable to that produced from a diffraction-limited PSF. However, because the new OTF has no regions of zeros,
digital processing can be used to restore the sampled intermediate image. If the PSF is insensitive to misfocus, all values
of misfocus can be restored through a single, fixed, digital filter. This combined optical/digital system produces a PSF
that is comparable to that of the diffraction limited PSF, but over a far larger region of focus. The phase cubic
distribution necessary to produce this focus-invariance is described by equation 4:
[ (
P( x p , y p ) = exp j20π x p + y p
3
3
)]
(4)
When a focusing error is present, the center of curvature of the spherical wavefront converges towards the image of an
object point source at the left or at the right of the image plane located at the distance zi from the exit pupil 8. The
misfocus in the generalized pupil function 3(xp,yp) for an image located at the arbitrary distance za from the exit pupil
results in the equation:
(
 1 1
1 2
2
−  x p + y p
P ( x p , y p ) = P( x p , y p ) exp  jk 
 2  z a z i 
)
(5)

The amplitude impulse response h(xi,yi) of this new lens is calculated by the equation 6:
+∞
h (x i , y i ) =
∫ ∫ exp[j20π(x
−∞
3
p
+ yp
3
)]exp jk 12  z1


a
−
1
zi
(
)

 2
 x p + y p 2  exp − j2π x p x i + y p y i dx i dy i (6)


[
(
)]
DE
Figure 4: (a) Image of the PSF for zi=za=50 mm; (b) Image for zi=50 mm and za=70 mm. The cubic phase distribution generates a
constant PSF over a far larger region of focus.
Figure 4 compares the image of the PSF for zi=za=50 mm and for zi=50 mm and za=70 mm. The cubic phase
distribution generates a constant PSF over a far larger region of focus. Figure 5 compares the numerical results of
misfocus imaging using the cubic phase distribution for zi=za=50 mm and for zi=50 mm and za=70 mm. The intensity
distribution of these intermediate images remains constant for different image planes. These results are calculated using
the relation:
G i ( u , v ) = H ( u , v )G g ( u , v )
(7)
DE
Figure 5: Numerical results of misfocus imaging using the cubic phase distribution. (a) image for za=50 mm (image plane); and (b)
image za=70 mm. Digital processing can be used to restore the original letter A from the sampled intermediate cubic phase image.
The OTF +(u,v) of the system is calculated using equation 2. The frequency spectra of Ig, the input image, and Ii, the
output image are defined by:
+∞
∫ ∫I
[− j2π(x
g ( x g , y g ) exp
gu
)]
+ y g v dx g dy g
−∞
G g ( u , v) =
(8)
+∞
∫ ∫I
g ( x g , y g )dx g dy g
−∞
+∞
∫ ∫ I (x , y ) exp[− j2π(x u + y v)]dx dy
i
G i ( u , v) =
i
i
i
i
−∞
i
i
(9)
+∞
∫ ∫ I (x , y )dx dy
i
i
i
i
i
−∞
As shown in figure 5, cubic phase optical systems are insensitive to misfocus. Because the new OTF has no regions of
zeros, the original images can be recovered from the focus invariant images after simple digital filtering using equation
10, as desired 2.
G g ( u , v) = G i ( u , v) / H ( u , v)
(10)
$33/,&$7,212)&8%,&3+$6(',675,%87,21,13$66,9(,1)5$5('027,21
6(16256
As stated above, in PIR sensors the lenses are used to concentrate the infrared radiation over the detector and not to
generate an image. Regular lenses do not distinguish the presence of humans from pets. In order to assure this
distinction, the cubic phase distribution is added to the regular lenses, introducing the wavefront coding in the system.
In our application, the new lenses are used as a spatial filter rather than as an imaging system.
When a target mass is moving through the volumetric monitoring field of the sensor, i.e., through the fileld of view of
different segmented Fresnel lenses, the generated image also moves over the detector surface. The system acts as a
horizontal one-dimensional imaging system as shown in figure 2. As stated above, we want to exploit any possible
frequency filtering introduced by the cubic phase distribution on the lens. The goal is to introduce a spatial filtering that
helps in the distinction between the patterns generated by a human and a pet body. To verify the best behavior of the
cubic phase mask in one-dimensional systems, this distribution is tested independently in the vertical direction, using
the PV(xp,yp) phase distribution, and in the horizontal direction, using PH(xp,yp), shown in the equations bellow:
[
]
, y ) = exp[j20πx ]
PV ( x p , y p ) = exp j20πy p
3
PH ( x p
3
p
p
(11)
(12)
Figure 6: Input test image used in the computer simulation. It contains several qualitative human bodies in different conditions,
rectangles, a pet and small squares simulating small pets.
Figure 6 shows the input test image used in the computer simulation. It contains several qualitative human bodies in
different conditions, rectangles, a pet and small squares simulating small pets. Figure 7 shows the output images of the
phase distributions of equations 11 and 12 multiplied by the phase distribution of a conventional lens, also shown in
figure 7. These lenses act as a one dimensional low-pass filter. The modified lens of figure 7a enhances the information
of horizontal features, and the lens of figure 7b enhances the information of vertical features, as well. One can note from
figure 7b that all intermediate images of the qualitative human bodies have higher bright images if compared to the
images of all other features. The lens of figure 7b turns possible to distinguish the presence of humans from pets.
DE
Figure 7: Output images of an one-dimentional cubic phase distribution multiplied by the phase distribution of a conventional lens.
These lenses act as a one dimensional low-pass filter. The modified lens of figure 7a, generated using the hozontal phase distribution
PH(xp,yp), enhances the information of horizontal features, and the lens of figure 7b, generated using the vertical phase distribution
PV(xp,yp), enhances the information of vertical features, as well. One can note from figure 7b that all qualitative human bodies have
higher brigth images if compared to the images of all other features. The lens of figure 7b turn possible the distinguish the presence
of humans from pets.
&21&/86,216
In this work we describe the wavefront coding method applied to passive infrared motion sensors. The new proposed
lenses have the ability to distinguish the presence of humans from pets. The lenses are designed considering a cubic
phase distribution. The cubic phase distribution generates an invariant intermediate image over a far larger region of
focus. This aberration control allows a low tolerance in the fabrication of the lenses and in the alignment errors of the
sensor. We verified the behavior of the cubic phase mask in one-dimensional systems, testing this distribution
independently in the vertical and horizontal directions. We simulated several qualitative human bodies in different
conditions, as well as other features, such as small pets. As a result, we demonstrate that the proposed system is able to
enhance the information of horizontal features or vertical features, turning possible, in this way, the distinction of the
presence of humans from pets.
5()(5(1&(6
1.
US patents: 4.052.616 (1977), 4.321.594 (1982), 4.535.240 (1985).
2.
Edward R. Dowski, Jr., and W. Thomas Cathey “Extended depth of field through wave-front coding” $SSOLHG
2SWLFV, , 1859-1866, 1995.
3.
Dowski, E. R. Jr. and Johnson, G. "Marrying optics & electronics: Aspheric optical components improve depth
of field for imaging systems", 2(0DJD]LQH, , 42-43, 2002.
4.
US patents: 5.748.371 (1998), 6.069.738 (2000).
5.
Dowski, E. R. Jr.; Cathey, W.T.; van der Gracht, J. “Aberration Invariant Optical/Digital Incoherent optical
Systems”, Publication of the Imaging System Laboratory Department of Electrical Engineering, University of
Colorado, Boulder, available (18/07/02) at: http://www.colorado.edu/isl/papers/aberrinv.pdf
6.
Cathey, W. T.; Dowski, E. R.; and Alan R. FitzGerrell, A. R. Optical/Digital Aberration Control in
Incoherent Optical Systems" Publication of the Imaging System Laboratory, Department of Electrical
Engineering,
University
of
Colorado,
Boulder,
available
(18/07/02)
at:
http://www.colorado.edu/isl/papers/aberration/paper.html
7.
Wach, H.B., Cathey, W. T.and Dowski, E. R “Control of Chromatic Focal Shift Through Wavefront Coding”
Publication of the Imaging System Laboratory Department of Electrical Engineering, University of Colorado,
Boulder, available (18/07/02) at: http://www.colorado.edu/isl/papers/chromaberr.pdf
8. Joseph W. Goodman. “,QWURGXFWLRQ WR )RXULHU 2SWLFV”, pp.137, McGraw-Hill Publishing Company, Second
Edition, New York, 1996.