You are on page 1of 7

366 J. Opt. Soc. Am./Vol. 73, No.

3/March 1983
Projection moire for remote contour analysis
J. L. Doty
Spectron Development Laboratories, Incorporated, 3303 Harbor Boulevard, Costa Mesa, California 92626
Received August 6, 1982
Remote projection and viewing of moir6 contours are examined analytically for a system employing separate pro-
jection and viewing optics, with specific attention paid to the practical limitations imposed by the optical systems.
It is found that planar contours are possible only when the optics are telecentric (exit pupil at infinity) but that the
requirement for spatial separability of the contour fringes from extraneous fringes is independent of the specific
optics and is a function only of the angle separating the two optic axes. In the nontelecentric case, the contour sep-
aration near the object is unchanged from that of the telecentric case, although the contours are distorted into low-
eccentricity (near-circular) ellipses. Furthermore, the minimum contour spacing is directly related to the depth
of focus through the resolution of the optics.
INTRODUCTION
The phenomenon of moir6 fringes was first described by Lord
Rayleigh in 1874.' He noted that when two matched line
gratings were placed in contact ". . . in such manner that the
lines are nearly parallel. . ." an additional series of wide par-
allel bars developed with characteristics that were a function
of the line spacing and inclination. However, the moir6
phenomenon was little used until recently because of the
difficulties encountered in the manufacture and reproduction
of satisfactory gratings. But in the early 1950's a novel
technique was developed
2
whereby diffraction gratings could
be reliably reproduced from a turned master grating, and the
field has since blossomed into a powerful metrological tech-
nique.
The application of basic moir6 contouring was first reported
in 1970 by Meadows et al.
3
Their technique was a near-
contact method in which the shadow of a grid was cast directly
onto a model surface by a collimated beam of incoherent ra-
diation. The grating spacing was large enough, and its dis-
tance from the model small enough, that diffraction effects
in the projected shadow were negligible. The model was
viewed through the same grating used to cast the shadow
fringe pattern, and the observer was located far enough from
the grid that the lines of sight from the observer to the grid
were essentially parallel for all points on the grid.
The result of such a configuration is that the model appears
to have not only the grid shadows but a series of additional
shadows (moir6 fringes), which represent the intersection of
equally spaced contour planes with the surface of the model.
The contour planes are literally a separate spatial frequency
generated by beating the spatial frequency of the illumination
grid with that of the observation grid. In this case both grids
are one and the same, and the contours are planes that are
parallel to the plane of the grid and have a spacing of
tan 0 + tan ck
where p is the grid spacing and 0 and 0 are the illumination
and observation angles, respectively. If the contour spacing
(Az) is much greater than the grid spacing (p), then the moir6
contours are easily separable from the grid shadows.
This form of basic moir6 contouring is extremely powerful
under the right circumstances. However, the grid must be
located relatively close to the model to prevent deleterious
diffraction effects, and it must be as large as or larger than the
model if full coverage is desired. It is these difficulties that
necessitate a modification of the basic moir6 technique for
remote and/or large objects.
Several authors have analyzed remote projection and
viewing of moire contours.>
7
Their analyses, however, have
concentrated on the details of contour shape and localization,
with only limited regard paid to the specific optics. The
purpose of this paper is to provide a detailed examination of
the projection moir6 technique, with specific attention paid
to the practical limitations imposed by (and on) the projection
and viewing systems.
PROJECTION MOIRt FOR REMOTE SENSING
The optical configuration of a projection moir6 system is il-
lustrated in Fig. 1. It consists mainly of a projection and an
observation arm that, respectively, project and view moire
shadow fringes on the model. In this analysis the model is
centered on the origin of an x, y, z Cartesian-coordinate sys-
tem. Note that the x axis is perpendicular to the plane of the
figure. Both the projection and observation arms have optic
axes that lie in the y, z plane, intersect the origin, and are re-
moved from each other by an angle 20. Note that the z axis
bisects this angle and that there is no loss of generality here
since unequal projection and observation angles can be rep-
resented by a simple rotational transformation of the model
coordinate space.
To produce contours on the model, an incoherent, white-
light source illuminates the grating in the projection arm,
which is in turn imaged onto the model surface. Now,
whereas the grating itself is two dimensional and has no sig-
nificant longitudinal depth, its image has a longitudinal depth
that is equal to the depth of focus of the imaging optics, which
is chosen to be equal to the maximum depth of the model.
Hence the image of the projection grating in the vicinity of the
model consists of dark and light planes that can be, but are not
necessarily, parallel to one another. These planes are illus-
0030-3941/83/030366-07$01.00 1983 Optical Society of America
J. L. Doty
Vol. 73, No. 3/March 1983/J. Opt. Soc. Am. 367
server is
1. (a, 3, y) = T. (a, 3, y)R (a, 1, y). (7)
Now since the imaging transformation of a properly cor-
rected imaging lens is linear, the transformation of Eq. (5) is
merely
R(a, (, y) = S(a, 1, y)I, (a, 1, y). (8)
Therefore, it follows from Eqs. (3), (4), (7), and (8) that
I.(a, 13, y) = IS((a, 1, y)T.(a, 1, y )Tp(a, 1, y). (9)
To interpret Eq. (9), which represents the observed image,
note that IsS(a, 1, y) is merely the image that would be re-
corded if both gratings were removed so that
To = Tp = 1. (10)
Fig. 1. Optical system for projection moire contouring.
trated by the shaded bands in the model space of Fig. 1. Note
that for clarity of the illustration only some of the grating lines
have been shaded in this fashion.
If the transmission function of the projection grating is
defined as Tp(a, b, c), where a, b, c is the Cartesian-coordinate
space of the projection grating, and if it is illuminated by the
uniform intensity Is, then the intensity seen by the imaging
lens is
I,(a, b, c) = IT,(a, b, c). (2)
The intensity distribution projected into the model coordinate
space is
where
(3)
(4)
represents the imaging transformation from the projection
grating to the model coordinate space. Note that imaging-
system effects on the uniformity of I, are ignored here since
they will have no effect on the final outcome of this anal-
ysis.
At the model, the intensity of the light scattered to the
observer is
R(x, y, z) = S(x, y, z)Ip(x, y, z), (5)
where S(x, y, z) is the localized point-scattering function of
the surface of the model, assumed to be diffuse and Lamber-
tian.
Now the imaging lens of the observation arm collects and
images this scattered light onto the observation grating. The
intensity distribution of this image before it passes through
the observation grating is R (a, 3, ry), where a, 1, -y is the
Cartesian-coordinate space of the observation grating and
R(x, y, z)=R(a, ,y) (6)
represents the imaging transformation from the model to the
observation-grating coordinate space. If the intensity-
transmission function of the observation grating is defined as
T, (a, 13, Y), then the intensity distribution seen by the ob-
However, with the gratings in the system, the intensity dis-
tribution of the model image is modulated by the combined
transmission function
T(a, 1, y) = T,(a, 1, y)T,(a, 1, y). (11)
The moir6 fringes are beat spatial frequencies that arise as a
result of the multiplication on the right-hand side of Eq.
(11).
To understand fully the manner in which this occurs, it will
be necessary to consider some specific grating-transmission
functions. However, when doing so, it will be more instructive
if the observed intensity distribution in Eq. (9) is referred to
the coordinate space of the model. This merely requires the
reverse of the imaging transformation in Eq. (6), which is itself
an imaging transformation from the observation grating to
the model coordinate space. And again, since such a trans-
formation is linear,
I. (x, y, z) IS (x, y, z) T(x, y, z), (12)
where
T(x, y, z) = T.(x, y, z)Tp(x, y, z). (13)
Note that, because of the simple linear transformations in-
volved, the moire phenomenon is equivalent to projecting the
shadows of both gratings onto the surface of the model in such
a way that the resultant intensity is the product of the two
projection intensities, rather than the sum, as would be the
actual case. This is perhaps the simplest way of under-
standing and analyzing the moir6 phenomenon, and it is the
method used throughout this paper.
FRINGE PROJECTION
The optical configuration for projecting the shadow of a
grating onto the model surface is illustrated in Fig. 2. For
clarity, the illumination source and condensing optics have
not been included, and the projection lens is modeled as a
simple lens with a remote aperture, which is accurate for a
well-corrected system.
The illumination is incident from the left. It strikes the
grating, which acts as a transmission filter, then passes
through the aperture stop to the lens. The object and image
positions (L and L') are related to each other through the
well-known relationship
J. L. Doty
I, (x, Y, Z) = I, TP (x, Y, Z),
I, (a, b, c) - I, (x, y, z)
368 J. Opt. Soc. Am./Vol. 73, No. 3/March 1983
Exit Pupil
V b Gr Aperture Stop
K Lap4 ag
L FF - 2
Fig. 2. Optics of fringe projection and viewing.
L L' F
where F is the focal length of the lens and the magnification
of the system is
m = L'/L. (15)
Therefore, if the grating has a line spacing of p, then its image
has a line spacing of p', where
D, = 2.44X(f#),
e = 2.44X(f#)
2
,
(21)
(22)
where X is the wavelength of the radiation and f# is the rela-
tive aperture in the image space of Fig. 2.
Se is therefore the range over which the grating shadows
are in focus in the image space of Fig. 2. Furthermore, to
ensure that two adjacent shadow planes in the grating image
can be resolved one from the other, the condition
p' 2 D, (23)
is imposed on the system.
Returning to the projected transmission function of Eqs.
(19) and (20), a rotational transformation is required to relate
it to the x, y, z coordinate space of the model. Figure 3 il-
lustrates the relationship between the model coordinate space
and the image coordinate spaces of the projection arm (a', b',
c') and the observation arm (a', 3', y'). To transform the
transmission function of Eqs. (19) and (20), the following di-
rect substitution is made:
a' =-z cos 0 - y sin 6,
V = +y cos 0 - z sin 0.
Hence
P, = mp.
(16)
Now the cone of rays that forms any point on the image ap-
pears to come from the exit pupil of the projection system.
The positions of the aperture stop and the exit pupil are re-
lated through
Lap Lex F (17)
which is a modified form of the imaging relationship in Eq.
(14).
Now consider a sinusoidal transmission grating with a
transmission function
Tp(a, b, c)= I + Cos (
2
w )1, (18)
where p is the line spacing of the grating and a, b, c is its
Cartesian-coordinate space as defined in Fig. 2. The image
of this grating is a fan of planar shadows in the a', b', c' coor-
dinate space (also defined in Fig. 2) that intersect at the center
of the exit pupil. The imaging transformation yields
where
T(x, Z) 1 + cos L27 Lp(Y, z)I}
L, (y, z) =(L' + Lex)(y cos 0 - z sin 0)
L'+Lex - z cos a -y sin 0
(24)
(25)
(26)
As before, the above shadow distribution is in focus only
within the depth of focus (E) and hence is subject to the con-
dition
Ia'I 5 e. (27)
The projected shadow planes of expressions (25)-(27) are il-
lustrated in Fig. 4.
TELECENTRIC PROJECTION MOIRE
CONTOURING
In standard projection moir6 contouring the objective is to
generate a contour map of the surface of the model. To do so,
Tp(a', b', c') = 2 [1 + cos [r LLp(a', b')I1
2 (. pfII
where
L (a', b') = (L' + Lex)b'
L' + L,,, + a'
Ia'l < e,
Projection Arm
Optic Axis
(19)
(20)
and e is the depth of focus of the optical system, which de-
termines the effective longitudinal depth of the planar
shadows.
The depth of focus (E) is here defined as half of the longi-
tudinal range over which the diameter of the geometric ray
spot is less than or equal to the diameter (D,) of the central
disk in the Airy distribution. Therefore
Observation Arm
Optic Axis
Fig. 3. Geometric relationship between the model coordinate space
(x, y, z) and the image coordinate spaces of the projection (a', b', c')
and observation (a', I', -y') arms.
J. L. Doty
Vol. 73, No. 3/March 1983/J. Opt. Soc. Am. 369
Inserting Eqs. (28) and (30) into Eq. (13) yields
T(x, y, z) = [1 + cos (7, Lp)]4[ + cos (, L
0
)], (32)
and, carrying out the multiplication,
T(x, y, z) = 1 [1 + cos ( , Lp) + cos ( ,L)
+ cos ( Lp) cos (, Lo)] - (33)
The last term in Eq. (33) can be expanded to yield
T(x, y, z) = 4 f1 + cos (2, Lp) + cos ( -X L,)
Exit
Pupil
Projection/
optic
Axis /
Fig. 4. Fringes cast by the projection arm.
one uses a straight-line grating with a transmission function
like that of Eq. (18) in both the projection and the observation
arms of the system. Note that most practical gratings exhibit
a square-wave transmission function as opposed to the si-
nusoidal distribution considered here. The sinusoidal anal-
ysis is, however, simpler and has been shown to yield the same
result with respect to contour location and spacing.
3
The projected shadow distribution of the projection arm
is taken to be
(28)
(29)
which is identical to Eqs. (25) and (26) but with the distance
from the model to the exit pupil (L' + Lex) replaced by the
variable dp for simplicity. As was stated before, the obser-
vation arm can be looked on as also casting shadow planes in
a fashion identical with that described in Eq. (28). But the
grating coordinate space of the observation arm corresponding
to Fig. 2 is a, 3, oy rather than a, b, c, and the grating image
space is a', f3', y' rather than a', b', c'. The reason for this
distinction is illustrated in Fig. 3, for which the final rotational
transformation from the a', ', ry' space to the x, y, z space is
carried out by using Eqs. (24) but with 0 replaced by -6.
Hence the projected shadow distribution of the observation
arm is
(30)
where
do (y cos 0 + z sin 6)
L.(yz) = c (31)
do, - zcos 0+ ysin 0
o' is the magnified line spacing of the shadow planes, and do
is the distance from the model to the exit pupil of the obser-
vation arm.
+ 2cos [27r ( + L-.)] + 2cos [2wr (- -11.
(34)
Now, if we remember that T(x, y, z) is the spatial modulation
of the intensity distribution of the observed image, any term
of the form
cos[2wf(x, y, z)] (35)
within T(x, y, z) describes a series of fringes whose location
is determined by
f(x,y,z) = N, (36)
where N is an integer and is here called the fringe number.
Therefore the intensity-modulation function of Eq. (34)
produces four sets of fringes determined by
LP = N. (37)
p
1
=, - N (38)
0'
P+' =N,
p/ - N
LP -L =N
LI L
0
(39)
(40)
where Lp and L, are defined in Eqs. (29) and (31).
Fringe Eq. (37) represents the original shadow planes pro-
jected by the projection arm and illustrated in Fig. 4. Fringe
Eq. (38) represents another set of shadow planes that appear
to be projected onto the model surface by the observation arm
but in actuality exist only behind the observation grating.
They, like the fringes depicted in Fig. 4, are a fan of planes.
But they differ from the fringes of Fig. 4 in that they fan out
from the exit pupil of the observation arm on the right-hand
side of the z axis.
Fringe Eqs. (39) and (40) represent the sum and difference
beat spatial frequencies of most interest here. To simplify
their characteristics, the following assumptions are made:
dp = do = d,
0 = I
(41)
and, with some manipulation, Eqs. (39) and (40) take the
form
y cos O - z sin H y cos 0 + z sin o Np'
* d-zcosO
= , (42)
d - zcos 0- ysin 0 d - zcos0+ y sin60 d
where
Tp(xyz) = [1 + cos [ , (y Z)
- dp(y cos O - z sin 0)
dp - z cos 0 - y sin 0
To(x, Y, Z) {= I + cos | Lo(Yz)11z
J. L. Doty
370 J. Opt. Soc. Am./Vol. 73, No. 3/March 1983
lA y
Fig. 5. Planar sum and difference fringes of the telecentric config-
uration.
where the + term represents the sum spatial frequencies [Eq.
(39)] and the - term represents the difference spatial
frequencies [Eq. (40)].
For contouring, the fringe shape is ideal if the aperture stop
is located in the front focal plane of the projection lens in both
arms of the system (Lap = F in Fig. 2). This is the telecentric
configuration, and its effect is to place the exit pupil at in-
finity, i.e.,
d = c. (43)
Equation (42) then simplifies to
(y cos 0 - z sin O) (y cos a + z sin G) = Np', (44)
which describes two sets of planar fringes in addition to those
of the projection fringes of Eqs. (37) and (38). As is illustrated
in Fig. 5, the sum fringes are parallel to the x, z plane with a
spacing of
Ay I
AY=2 cos 0,
y = 0,
z = d cos 0.
(49)
At N = X, the two hyperbolas again collapse to form two
straight lines defined by the equations
z = - + y tan 0,
cos 0
d
Z = - y tan 0.
cos 0
(50)
Note that the asymptotes of the hyperbolas of a given fringe
order (N) are not perpendicular and that, as N varies between
the limits of -tX, the origin of the y', z' system (yo, zo) traces
out an ellipse in the y, z plane and i varies from -45 to
+450.
Of primary importance, though, is the spacing of the fringes
near the origin of the model space (y, z). Expanding Eq. (48)
and ignoring all second-order terms yields
Np'(d_
Y d 2cos .Z
(51)
(45)
and the difference fringes are parallel to the x, y plane with
a spacing of
If 0 is small so that
then the sum fringes as well as the two sets of projection
fringes have a spacing of approximately p', whereas the
spacing of the difference fringes is much greater. The dif-
ference fringes, therefore, appear as spatially separable con-
tour planes parallel to the x, y plane of the model coordinate
space.
LARGE OBJECT PROJECTION MOIRE
CONTOURING
Unfortunately, the telecentric configuration requires optics
as large as or larger than the model, which makes it, like basic
moir6, impractical for large objects. For that reason, the
NcO
Fig. 6. Locations of the hyperbolic Sum fringes with varying fringe
order (N).
more-general case of a finite exit-pupil location is considered
next by dealing with Eq. (42) without further simplifica-
tion.
For the case of the sum fringes, Eq. (42), after further ma-
nipulation, yields
2y(d cosO - z) = NP Rd -z COS 0)
2
- y
2
sin
2
0]. (48)
d
There is no simple or convenient form for Eq. (48). However,
by expanding to the general form and comparing with known
cases, it is found that Eq. (48) represents a complex series of
hyperbolas. Their locations are best understood by referring
them to ay', z', coordinate space that has an origin located at
yo, zo in the y, z model space and has been rotated through an
angle a7, as illustrated in Fig. 6.
At N = 0, the two hyperbolas collapse to form to straight
lines defined by the equations
Aa
2 sin 0
sin 0 << cos 0,
(46)
(47)
J. L. Doty
____ -__J
Az
__7
--------
Y .4
-------
--------
Vol. 73, No. 3/March 1983/J. Opt. Soc. Am. 371
singularities. Hence Eq. (55) takes the form of a series of el-
lipses for B
2
positive and a series of hyperbolas for B
2
negative.
Again, however, the fringes of most importance are those near
the origin where the model is located, which correspond to
fringe-order numbers near N = 0.
Figure 9 illustrates the location of the elliptical difference
fringes for the range
tan 6 - cot 0 <A < 2 tan 6. (61)
Fig. 7. Approximation to the hyperbolic sum fringes in the imme-
diate neighborhood of the model.
In the neighborhood of the model, then, the sum fringes can
be approximated by a fan of planes with a spacing of
Y 2 cos 67 (52)
This is identical with the spacing of the telecentric case, but
the fringes are no longer parallel to the x, y plane. Rather,
they appear to radiate from the coordinates
y =0,
d
2 cos 6 (53)
as illustrated in Fig. 7.
Next, for the case of the difference fringes, Eq. (42), after
rearranging, yields
2 sin 0[y
2
cos 0 - z(d-zcos0)]
= Np [(d - z cos 0)2 - (y sin 0)2],
d
which, after further manipulation, becomes
where
(Z - Z.)
2
y2
-+-= 1,
A
2
B
2
(d --coto
Z = 2cosO 1-Acot6),
d 12
A2 -2 cos 6)2
A =
(1 -~cot6)
2 8
B
2
= (2c s o)2
(1-A cot o) ( + Atan 0)
Np'
1 d
(54)
(55)
(56)
(57)
(58)
(59)
Figure 8 is a normalized graph of both A
2
and B
2
as functions
of the fringe-order parameter ,u. Both have singularities at
A = -2 cot 0, ,u = +2 tan 0. (60)
But, whereas A
2
is always positive, B
2
changes sign at both
The lower limit in expression (61) was chosen to be the value
at which B
2
is a minimum. Note that as It increases from the
lower limit the fringes form expanding ellipses until the upper
limit is reached, at which point they open out into a parabola.
Beyond this they are again hyperbolas. Also note that
A
2
= B
2
, A =
0
.
(62)
Therefore near the origin the fringes are nearly circular, with
a spacing of
2 sin 0 (63)
As with the sum fringes, the difference fringes in the
neighborhood of the model are nonplanar, and their spacing
is identical with that of the telecentric case. And since the
fringe spacing near the origin of both the sum and difference
fringes is unchanged from that of the telecentric case, the re-
quirement for spatial separability [expression (47)] is also
unchanged.
N-2 cot6 2 tanG,--
;i 'B2
tanG-cotG If
Fig. 8. Variation of the major and minor axes of the elliptical dif-
ference fringes as functions of the fringe-order parameter ,u.
Fig. 9. Location of the elliptical difference fringes in the neighbor-
hood of the model.
J. L. Doty
372 J. Opt. Soc. Am./Vol. 73, No. 3/March 1983
CONCLUSIONS
It is possible, and practical, to project moir6 contours onto
remote objects, but matched, physically separate projection
and observation systems are required. Furthermore, to be
able to separate the contour fringes spatially from extraneous
fringes, the angle between projection and observation arms
must be kept small [see expression (47)]. Limitations on the
contour spacing [expression (23)] and applicable depth [ex-
pression (27)] are imposed by the projection and viewing op-
tics.
Planar contours are possible only if a telecentric optical
configuration is used in both arms of such a system. But, like
simple near-contact moire, the telecentric configuration limits
the concept to contour objects no larger than the optics em-
ployed. To work with large objects the user must resort to a
nontelecentric configuration, which distorts the shape of the
contours near the object into low-eccentricity ellipses.
However, the contour shape is readily predictable, and the
contour spacing near the object and the requirement for
spatial separability are both unchanged from that of the
telecentric configuration.
If one makes the simple assumption that contours are not
visible outside the depth of focus of the system, the total
number of usable contours generated is
Nj = A cos 9, (64)
where E and Az are taken from Eqs. (22) and (46). Hence,
after manipulation,
N = 2 (f#) sin 20. (65)
Furthermore, by making use of the restriction in expression
(23), the maximum number of contours that it is possible to
generate is
Nt ' 2(f#)sin 20. (66)
Ultimately, this is the limit of the contour resolution that
a projection moire system can achieve.
ACKNOWLEDGMENTS
My sincere appreciation goes to B. P. Hildebrand for his
guidance and supervision throughout this program. This
research was funded under contract to the NASA Langley
Research Center (NASA contract no. NAS1-16564).
REFERENCES
1. Lord Rayleigh, "On the manufacture and theory of diffraction-
gratings," Philos. Mag. 47, 81-93, 193-205 (1874).
2. T. Merton, "Nouvelles methodes de fabrication des reseaux," J.
Phys. Radium, 13, 49-53 (1952).
3. D. M. Meadows, W. 0. Johnson, and J. B. Allen, "Generation of
surface contours by moire patterns," Appl. Opt. 9, 942-947
(1970).
4. M. Idesawa, T. Yatagai, and T. Soma, "Scanning moire method
and automatic measurement of 3-D shapes," Appl. Opt. 16,
2152-2162 (1977).
5. L. Pirodda, "Shadow and projection moir6 techniques for absolute
or relative mapping of surface shapes," Opt. Eng. 21, 640-649
(1982).
6. T. Yoshizawa and H. Tashiro, "Localization of fringes in moire
topography," Opt. Lasers Eng. 3, 29-43 (1982).
7. M. Takeda, "Fringe formula for projection type moire topography,"
Opt. Lasers Eng. 3, 45-52 (1982).
J. L. Doty

You might also like