International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004
In a projection moiré, a reference grid is projected onto the
object. The projection is distorted by the object topography.
This distorted version of the grid is imaged through the
reference grid, and a moiré pattern is formed on the image
plane.
The main disadvantage of the methodology is that the depth of
the surface under test is limited to not more than several tens of
the wavelengths of light used. (Yu et al. 1997; Paakkari et al.
1998; Mikhail et al. 2001; Karara 1989)
3.2 Light Stripping
Light stripping is a particular case of the structured light
technique for range sensing. A light striping range sensor
consists of a line projector and a camera that is displaced from
the light source. The line projector casts a sheet of light onto
the scene. The projection of the plane on the surface results in
a line of light. This line is sensed by the camera. The light
plane itself has known position in world coordinates that can
be determined by measuring the geometry of the projection
device or by direct measurements of the resulting plane of
light. Every point in the image plane determines a single line
of sight in three-dimensional space upon which the world point
that produces the image point must lie.
This methodology requires a large amount of images. In this
method each image contributes a single profile of the object.
Generating a DEM with the desired resolution and speed
requires capturing and processing of 300 frames per second
(Mikhail et al. 2001; Karara 1989).
3.3 Coded Light Projection Techniques
Among all the ranging techniques, stereovision is based on
imaging the scene from two or more known points of view and
then finding pixel correspondences between the different
images in order to triangulate the 3D position. Triangulation is
possible if cameras are calibrated (Salvi 2001). However,
finding the correspondences poses difficulties even after taking
into account epipolar constraints. Coded structured light
consists of substituting one of the cameras by a device that
projects a light pattern onto the surface. Since each point in the
pattern is encoded in a way that identifies its coordinates, the
correspondence problem is solved with no geometrical
constraints. With an appropriate light pattern, correspondences
of a large number of pixels can be found. Thus, coded
structured light may be used as a solution to simplify the
inherent problem of finding correspondences in classical
stereovision systems. (Pag'es 2004)
4. OBJECTIVE
Considering the advantages and disadvantages of the
methodologies presented above, the specific objective of the
present work was to develop a DEM generation system for
chicken filets using a coded light projection technique.
5. DESCRIPTION OF THE SYSTEM
The imaging system contains several major hardware parts
(Fig. 1) (a) a projection unit including the projector itself, (b) a
computer which creates patterns that are projected on the
object and (c) two cameras to capture a stereoscopic view of
the object. The images from the two cameras are sent to frame
grabber in the host computer where the analysis is preformed.
ertet errore di "
Figure 1 Major hardware parts
In order to calculate the object's height the first task is to
determine the point in the space, where the cameras center of
projection is located, and its orientation. This process is
essential in order to build the 3-D model of the object and for
finding the epipolar lines. The aim of this calibration process is
to find the six extrinsic parameters for each of the cameras.
These parameters contain three location parameters (Xt, Yt, Zt)
in millimeters, and three orientation parameters (phi, epsilon,
kappa) in radians. In order to calculate the parameters, an
image of a calibration object with marked points of known
world coordinates was acquired by each one of the cameras.
SPR (Single Photo Resection) method was used to estimate the
extrinsic parameters of each camera. (Mikhail et al. 2001)
After obtaining an estimate of the cameras' extrinsic
parameters, the height of any point on the object is calculated
by finding its corresponding pixels in the two images and the
principle of triangulation. In order to overcome the
correspondence problem a continuous color pattern is projected
on the object.
6. THE PROJECTED PATTERN
Several methods for overcoming the correspondence problem
using projected color patterns can be found in the literature
(Tajima et al. 1990; Geng 1996; Sato 1997; Wust and Capson
1991). Those researches present global methods to use a
projection that will fit any object. Those objects may be multi-
color and multi-texture bodies. Caspi et al. (1998) analyze the
illumination model of a structured light system. This model
takes into account the light spectrum of the LCD projector, the
spectral response of a 3-CCD camera and the surface
reflectance. The main benefit of this model is that it assumes a
constant reflectance for every scene point in the three RGB
channels (Pag'es 2004). Preliminary tests showed that when
transforming an image of a color projection captured in RGB
to HSL (Hue Saturation Lightness) the hue channel contains
most of the information, while the other two channels contain
mostly noise. Caspi's work also defined a noise immunity
factor a, and presented the relationship between a and the
number of possible strips. The number of possible strips in the
projected pattern depends on the maximum accuracy under
constrains of the projection / sensors equipment.
The structured light technique that is presented in the present
work can cover different values of a, across the different RGB
intensity values. The sensitivity of the CCD sensors changes
across the spectral range. The projected patterns can be
Internatio
squeezed
system ca
The color
among fil
system p
matching
uniform «
uniform
spectrum
in some 1
change di
the projec
Projectior
on the ol
pattern o
chromatic
projectior
that the 1
similar c
reflected
the captu
change d
systems s
one wave
Assuming
captured
specific «
develop t
that will
values. F:
and chro
pattern th
To find 1
implemer
with line:
pattern at
The s
1. & Proj
2. Capt
the c
3. Com
patte
diffe
Ci -
4. fal
€ ti
else
rate,
The metl
linearly v
board anc
result of
X-axis is
pattern (f
the first i
is the de
pattern c
according
0.7. In th:
shows th:
One can :
