Note: Descriptions are shown in the official language in which they were submitted.
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
COMBINED STEREOVISION, COLOR 3D DIGITIZING
AND MOTION CAPTURE SYSTEM
FIELD OF THE INVENTION
The present invention relates to a digitizer combining
functions of stereovision, color 3D digitizing and motion
capture of a target object, a digitizing system using the
digitizer, and associated digitizing and motion tracking
methods.
BACKGROUND
3D digitizing, particularly non-contact optical 3D
digitizing techniques, have become commercially available
during recent years. Most of these techniques are based on
the principle of optical triangulation. Despite the fact that
passive optical triangulation (stereovision) has been studied
and used for many years for photogrammetric measurements, the
active optical triangulation technique (particularly laser
scanning technique) has gained popularity because of its
robustness and simplicity to process the obtained data using
a computer. Most of the systems based on the active optical
triangulation principle were developed for industrial
applications, such as robotics assembly, robot guidance,
industrial inspection, reverse engineering, etc. A laser beam
or a laser stripe is projected onto a 3D surface of an
object, scattering the laser beam or laser stripe on the
surface. It is measured using a photo-electronic device. A
signal can be measured indicating the position (usually the
depth) of the measuring point. In most cases, the basic
measurements are either a point or a section profile. A
mechanical or optical scanning device is usually used to
provide a frame of 3D measurement. Laser is a monochromatic
light source that does not provide full color information.
1
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
So, an additional camera and light source are used when a
color texture is needed.
A new category of optical color 3D digitizers, such as
the present applicant's product line, has been developed.
These systems use structured white light projection combined
with a CCD camera allowing for the measurement of 3D geometry
and color texture of a surface. The projected structured
light (viewed by a camera from an angle different from the
light projection) is deformed due to the 3D surface relief.
The 3D coordinates of the surface are calculated by analyzing
the deformation. These kinds of systems are being used in
computer animation, special effects and in electronic game
development.
On the other hand, the passive optical triangulation
(stereovision, for example) is largely used for the purpose
of motion capture. The correspondence problem (automatically
finding one point on the object's surface from two optical
sensors, cameras in general) is not a major obstacle for this
application because only a limited number of points must be
measured. These points are often characterized by using
visible markers.
Another application of stereovision is stereoscopic 3D
display. Instead of determining the 3D coordinates of some
points of an object in a 3D space, it simply needs to display
a pair of stereoscopic images on a monitor (TV or computer
monitor) so that the 3D perspective of an image can be seen.
One possible configuration is to capture a pair of images
using two cameras, which observe the parallax effect of an
object. Then the left eye will view one image of this pair of
stereoscopic images and the right eye will view the other.
The human brain can easily merge this pair of images so that
the object is viewed as a 3D image.
The existing 3D digitizing systems and optical motion
capture systems are, in general, complex and too expensive
2
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
for the Internet and mass consumer applications. Most of
these systems incorporate sophisticated optical, electro-
optical, mechanical and electronic components. Special
expertise is needed to operate such a digitizer. In addition,
the existing systems support separately the 3D digitizing and
motion capture functions.
SUMMARY
An object of the invention is to provide a digitizer
combining functions of capturing stereoscopic images, color
3D digitizing, and motion capture.
Another object of the invention is to provide a system
using the digitizer, which is simple in construction, simple
to use and affordable for Internet and mass consumer
applications like conference via Internet, 3D Web, e-
commerce, off-line and on-line games and any application
which requires affordable 3D digitizing and/or motion capture
solution.
Another object of the invention is to provide methods
for digitizing and tracking motion of a target object, which
are implementable using a personal computer and simple
lighting and video camera components.
According to the present invention, there is provided a
digitizer combining functions of stereovision, color 3D
digitizing and motion capture of a target object, comprising:
a first camera;
a second camera;
a first projection arrangement having a first light
projector providing lighting for an active 3D range sensing
for each of the cameras, and a grating element in front of
the first light projector for projection of an encoded
pattern on a surface of the target object;
3
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
a second projection arrangement having a second light
projector providing lighting for an acquisition of texture
information of the target object;
a base onto which the first and second cameras and the
first and second projection arrangements are mounted in fixed
relative positions with respect to one another, the cameras
having optical axes converging through a single point, one of
the light projectors having an optical axis intersecting with
the optical axes of the cameras at the single point; and
a communication port connected to the cameras and the
light projectors, for reception of control signals setting
operation of the cameras and the light projectors and
transmission of video signals from the cameras.
According to the present invention, there is also
provided a digitizing system comprising a.digitizer as above
described and a computer having a port connectable with the
communication port of the digitizer, functions controlling
operation of the digitizer by generating the control signals
for the stereovision, color 3D digitizing and motion capture,
and functions for a processing of the video signals received
through the port and generation of digitized data as a result
of the processing.
According to the present invention, there is also
provided a method for digitizing a target object, comprising
steps of:
capturing basic images of the object with first and
second cameras without additional illumination of the object,
the cameras having optical axes converging through a single
point and being aligned in angled directions with respect to
each other so that the cameras have fields of view having
significant overlapping portions over a depth of measurement
including the single point;
illuminating the object with light in which an encoded
pattern is projected;
4
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
capturing structured images with the cameras;
illuminating the object with light deprived of a
pattern;
capturing texture images with the cameras;
identifying elements of the encoded pattern in the
structured images;
determining a position of the elements to produce a set
of measured points;
determining 3D coordinates of the measured points using
calibration information in respect with position and
alignment of the cameras;
determining coordinates corresponding to each measured
point in the texture images to produce a digitized image of
the object.
According to the present invention, there is also
provided a method for tracking motion of a target object,
comprising steps of:
capturing in parallel sequences of images of the target
object with first and second cameras having optical axes
converging through a single point and being aligned in angled
directions with respect to each other so that the cameras
have fields of view having significant overlapping portions
over a depth of measurement including the single point;
detecting control points in a first image of each
sequence;
tracking the control points in the sequences of images;
determining disparities between the control points in
the images from the first camera and the images from the
second camera;
determining 3D positions of the control points in
corresponding ones of the images taken at a same time by the
first and second cameras by using the disparities and
calibration information comprising relative position and
angular alignment of the cameras; and
5
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
generating trajectories of the control points as
sequences of the 3D positions of the control points
respectively.
Preferably, the system of the invention incorporates
elements to capture and transfer a pair of stereo images, to
obtain 3D coordinates and the color texture of a surface, and
to capture the displacement of a number of given points in a
real or quasi-real time. The data capturing process is
simplified to make the operation of the digitizer as
automatic as possible. With the system according to the
invention, the 3D model can be first created with some
identifiable control points located on the model surface.
Then the 3D position of these control points can be captured
in real or quasi-real time, so that the whole model can be
controlled or animated.
The cameras can observe disparity created by the active
light projector. There is also disparity in a pair of images
captured by the two cameras. This digitizer can be connected
to a computer via a digital port like a USB port, or other
standard high-speed connections. The computer controls the
cameras and respective light projectors. A snapshot button
can be used to launch a 3D measurement process. The digitizer
can be mounted onto a rotational table, with the computer
directly controlling the rotation of the rotational table. It
is also possible to place the target object on the rotational
table so that the angular position of the rotated object can
be known.
Preferably, the system provides at least three
functions.
First, a pair of cameras can capture a pair of
stereoscopic images at video rate. The stereo 3D image can be
created when these two images are displayed on a monitor,
which sends one image to the left eye and another image to
the right eye. These images can be transferred via a high-
6
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
speed link (Ethernet, Tl, T3, for example) to another
computer.
Second, combining the light projectors and both cameras,
the digitizer provides measurement of the 3D coordinates of a
surface with texture information. The encoded pattern is
projected on the surface of an object by a light projector
and both cameras capture the scene. With the surface relief
of the object, the projected pattern is deformed from the
point of view of the camera. With a careful calibration
technique, it is possible to determine the 3D coordinates of
some points on this surface by measuring the deformation of
the projected pattern. In principle, a combination of one
camera and one light projector can carry out the measurement
of the 3D coordinates. The use of two or more cameras, which
cover a common space, combined with one light projector,
provides three major advantages. First, the weighted average
values of the 3D measurements obtained by each of the cameras
correspond to a better 3D measurement. Second, this
configuration overcomes more problems caused by an occlusion
effect. Third, the two cameras observe the projected pattern
from different views so that a better interpretation of the
deformation of the projected pattern on a discontinued
surface can be obtained. The measured 3D coordinates can be
transferred via a high-speed link to another computer.
The third function is to make a motion capture of a
limited number of points in 3D space. When using some markers
on a surface, it becomes relatively easy to determine the 3D
positions of these points. It is also possible to use some
known points on the object, like the features of the skin,
lips, eyelids, eyes, etc. Of course, the 3D space observed by
the cameras must be calibrated and a disparity of a given
point captured by the cameras can be evaluated and its 3D
position can be calculated. When the number of points to be
measured is low, it is even possible to determine the 3D
7
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
positions of these points several times per second. This data
can be used to control the motion of an object or model. This
data can be transferred via a high-speed link to another
computer.
Motion tracking, which analyzes the dynamic motion of a
subject in a scene captured by any one or both cameras
mounted onto a rotational table, can be achieved with the
system according to the invention. A servo control may be
used to control (in real time) the rotation of the rotational
table in such a way that the digitizer can follow the dynamic
motion of the subject. A user, either present or at a
distance, can also send a command to the computer in order to
orient the digitizer to a desired direction. Since the servo
control provides the exact position of the rotational table,
it is evident that the whole 3D space covered by the
digitizer mounted on the rotational table is calibrated as a
known geometric space with respect to the digitizer. This
function provides the possibility to cover a larger space to
perform the three basic functions of this apparatus.
Instead of mounting the digitizer on the rotational
table, sometimes, it is convenient to place an object on this
table. This configuration simplifies the operation to merge
several views of a 3D object to create a complete 3D model. A
3D digitizer can measure the 3D surface of one single view of
an object. In order to create a complete object model, it is
necessary to capture different views of an object. When the
object is placed on the controlled rotational table, the
precise position of each view with respect to the other views
is known. So it is easy to register several views in a common
3D coordinate system and to merge them to create a complete
3D model.
8
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
BRIEF DESCRIPTION OF THE DRAWINGS
A detailed description.of preferred embodiments will be
given herein below with reference to the following drawings,
in which like numbers refer to like elements:
Figure 1 is a schematic diagram illustrating a possible
working environment of a digitizing system according to the
invention.
Figure 2A is a schematic diagram illustrating a front
view of a digitizer according to the invention.
Figure 2B is a schematic block diagram illustrating a
simplified inside view of a digitizer according to the
invention.
Figure 3 is a schematic diagram illustrating a
digitizing system with a digitizer mounted on a rotational
table 32 according to the invention.
Figure 4 is a schematic diagram illustrating a
digitizing system with a target object placed on a rotational
table according to the invention.
Figure 5 is a schematic diagram illustrating a possible
pattern used in a digitizer according to the invention.
Figure 6 is a schematic diagram illustrating a possible
control system of a digitizer according to the invention.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
Referring to Figure 2B, there is shown an embodiment of
a digitizer 1 combining functions of stereovision, color 3D
digitizing and motion capture of a target object, for example
an article 35 as shown in Figure 4 or a human subject 17 as
shown in Figure 1. The digitizer 1 has a first projection
arrangement formed of a first white light projector 3
optically coupled with a grating element 5 in front of the
projector 3 for projection of an encoded pattern in the light
provided preferably through a projection lens 4 and directed
9
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
on the target object. The digitizer 1 has a second projection
arrangement formed of a second white light projector 6
providing lighting without pattern preferably through another
projection lens 7. A first color camera 8 preferably with its
lens 9 and a second color camera 10 preferably with its lens
11 are arranged preferably next to the projection
arrangements. The light projectors 3, 6 can be operated using
respective lamp driver boards 12, 13 which are connected to a
main control board 14 with a peripheral interface 15, e.g. a
USB port. The main control board 14 and the cameras 8, 10 are
linked respectively to the port 15 so that signals are
communicated to and from a computer 2 by way of a cable 16.
The various components of the digitizer 1 are preferably
enclosed in a casing 18.
Referring to Figure 2A, the projection arrangements with
the projection lenses 4,7 are preferably positioned
vertically one above the other and aligned in angled
directions with respect to each other so that the two
projected light fields cover each other at a desired distance
of measurement. Likewise, the cameras with the projection
lenses 9, 11 are preferably positioned horizontally and
spaced from each other. As a result, the camera and
projection arrangements are positioned symmetrically with
each other in a cross-like arrangement.
Referring to Figure 2B, the grating element 5
controllably produces a fringe pattern 36 such as the one
illustrated in Figure 5, in the light projected by the
projection lens 4. The cameras 8, 10 are positioned aside
from the projection lenses 4 and 7 and aligned in angled
directions with respect to each other so that the cameras 8,
10 have fields of view 19, 20 (see Figure 1) directed on the
illuminated surface of the object and maximizing the
overlapping with each other over the depth of measurement of
the object. The video signals produced on the respective
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
video outputs of the cameras 8, 10 represent complementary
images of the object with a common image portion as a result
of the fields of view 19, 20 being overlapping. The cameras
8, 10 may be CCD cameras having respective video outputs that
are connected to the port 15 preferably accessible on a rear
side of the casing 18.
Referring to Figure 5, the encoded projected pattern 36
consists of a 2D graphic pattern. This pattern is a sequence
of color lines. The preferred pattern is a repeating sequence
of white, black, yellow, black, green, black, cyan and black.
The sequence is for example repeated 10 times. A magenta
reference fringe 37 replaces the last black line in the 5th
sequence. The orientation of the lines can be vertical,
horizontal or any desired angle. The width of the lines can
be constant everywhere on the pattern. They can also be
variable. The magenta line serves as an identifiable
reference fringe having an absolute 3D position determinable
in the images captured by the cameras 8, 10, using
experimentally predefined functions. The reference fringe 37
is preferably positioned so that each one of the cameras 8,
10 captures it over the entirety of the object 17, 35 (see
Figures 1 and 4).
Referring to Figure 2B, the optical axis of the
projection lens 4 should be perpendicular to the surface of
the projected pattern. The aperture of this lens 4 and its
focus should be well adjusted so that the projected pattern
will be sharp enough over all the depth of measurement.
The optical axes of camera lenses 9, 11 intersect with
the optical axes of the projection lenses 4, 7 at identical
distances from the apparatus. These four optical axes
intersect at the same point, which is usually referred as the
optical center of the apparatus. The angle between the
optical axis of the camera 8 and the optical axis of the
projection lens 4 determines the sensitivity of the distance
11
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
measurement, referred to as depth, in the direction of the
optical axis of the projection lens 4. The bigger the angle
is, the more sensitive the depth measurement will be. On the
other hand, the angle should not be too big, in order to keep
an overlapped area over the depth of measurement covered by
the two cameras 8, 10 at least equal to 80-90% of the field
of view 19, 20 of each camera 8, 10. As the projected pattern
36 is a repeating pattern of color lines, it is impossible to
differentiate between lines of the same color. Therefore the
measurement is without ambiguity only over a certain depth of
measurement. The bigger the angle between the optical axes of
the projection arrangement and the camera 8, 10 is, the
smaller is the depth of non-ambiguity.
Once the digitizer 1 is assembled, the optical and
mechanical configuration of the system is fixed. In order to
acquire 3D data, the system is calibrated. In the process of
calibration, the relationship between the position of the
elements of the projected pattern 36 and their distance to
the measurement system is determined.
Referring to Figure 6, the cameras 8, 10, the light
projectors 3, 6 and the main control board 14 of the
digitizer 1 can be powered by a power supply with, for
example, three outputs 21, 22, 23 providing appropriate
voltage levels, the power supply receiving energy through a
connector 24 preferably positioned on the rear side of the
casing 18.
Referring to Figures 1 and 6, a system using the
digitizer 1 according to the present invention is provided
with a computer 2 with a port 38 for acquiring the video
signals coming from the cameras 8, 10 and to send commands
from the computer 2 to the main control board 14. The
computer 2 will process the video signals coming from the
cameras 8, 10. The computer 2 will also initiate a
measurement sequence and control the two light projectors 3,
12
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
6 in a way that images with a projected pattern 36, without
pattern and with uniform white illumination are acquired. The
acquisition and projector control are preferably controlled
via the main control board 14 by the computer 2 through a
cable 16 connected to the port 15.
Preferably, the measurement process for the 3D
digitizing is as follows. Images are first acquired with the
cameras 8, 10 without illumination. The light projector 3 is
then energized in order to generate a pattern projected on
the object 17. Images are acquired with the cameras 8, 10
with the projected pattern. The light source of the
projection arrangement 3 is then deenergized and the light
source of the projection arrangement .6 is energized in order
to generate a uniform white light illumination of the
measurement space. Images are acquired with the cameras 8, 10
with uniform white illumination. Elements of the projected
pattern are identified in the images of the object to be
measured with the projected pattern. The position of the
elements of the projected pattern is determined in the images
of the object to be measured with the projected pattern. This
step generates a set of points measured. The 3D coordinates
of all measured points are determined using the calibration
information. The coordinates corresponding to each measured
point in the texture images acquired with the uniform white
light illumination are further determined.
For every recording sequence, a set of video images is
processed into a suitable software application to retrieve
the 3D information from the structured images, along with the
texture information. Different algorithms can be used for
retrieving the shape and texture information from the video
images using projected structured light. Most of these
algorithms are based on the identification and localization
of the fringes.
13
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
The first step of these algorithms will be to detect the
projected fringes in the acquired images. Image processing
algorithms are used to segment the fringes, which permits to
determine for each pixel of the image of which fringe the
pixel is a part of. Once the fringes are segmented, the
position of all fringes in all lines of the image can be
determined by finding the maximum intensity of the'fringe or
by calculating the center of gravity of the intensity
distribution of a fringe. If colored fringes are used, the
color of each fringe is determined.
The reference fringe 37 (see Figure 5) encoded into the
projected pattern is made uniquely identifiable, preferably
by its color. The next step in the 3D digitizirig process is
the segmentation and localization of the reference fringe 37
in the acquired images. For most lines, the position of the
reference fringe is then known. For each line of the image,
the fringes are numbered corresponding to their color
beginning from the reference fringe to the edges of the
image. The color permits to number the fringes correctly as
long as the object contains no steps that are larger than the
depth of non-ambiguity. The positions of the identified
fringes are then converted from real world coordinates into
real world units, like millimeters for example. For each
camera 8, 10 in conjunction with the light projector 3, a
calibration table is created in a separate calibration
process. This calibration table permits to convert the
measured positions of the identified fringes in 3D
coordinates in geometric units.
The conversion function can be:
(x, y, z) T= f(u, j, n)
where: (x,y,z)T are real world coordinates;
n is a number of the identified fringe;
14
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
j is a line of the image in which the fringe n is
identified; and
u is a position of fringe n in line j.
Once a point cloud is generated based on all the 3D
coordinate measurements from a set of images acquired by one
of the cameras 8, 10, for each point of the point cloud a
color texture vector is determined calculated from the
coordinates and the texture image acquired with uniform white
light illumination.
The computer 2 may be used to determine the color
detected in each pixel of the image obtained with white light
projection and, for each pixel, to determine the ratio of
each of the red, green and blue color channels to the maximum
of the three channels. Color enhancing can then be achieved
by applying the inverse of these ratios to the corresponding
pixels in the image obtained with the structured light
projection.
This process of 3D digitizing is executed for the camera
8 and the camera 10 generating two independent 3D models. In
a final step, these two models can be merged. The relative
position of the models is well known from the calibration
process so the merging of the models and textures can be
achieved by using an algorithm of cylindrical projection or
cubical projection.
A single full digitized image of the object is thus
obtained as a function of the spatial relationship between
the real 3D position of the detected fringe maxima in each
line of the image with structured light projection and the
real 3D position of the structured light projection, the
texture image being integrated into the digitized image.
Referring to Figure 4, when the object to be measured is
placed on a rotational table 32 connected to the computer 2
through a cable 34 for the purpose of capturing multiple
views, the computer 2 controls precisely the position of the
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
rotation so that the relative position between each view is
known. It is very straightforward to put multiple partial
models together to create a complete model.
Referring to Figure 6, the main control board 14
receives commands from and transmits signals to the computer
2 via the cable 16. In the illustrated embodiment, the main
control board 14 has a CPU 26, a USB interface 27, RAM and
PROM memories 28 for data and program and an Input/Output
unit 29. The commands that are recognized by the board are,
for example: resetting the digitizer 1; turning the light of
the light projector 3 on and off; turning the light of the
light projector 6 on and off; and starting the digitizing
process.
The lamp driver boards 12, 13 receive on/off signals via
the main control board 14. The intensity of the light
projector 3, 6 can be manually adjusted with potentiometers
30, 31.
Referring to Figures 1 and 2, to perform stereovision
display, the images captured by the two cameras 8 and 10 are
first transferred to the host computer 2, then a stereo pair
of images is calculated, the images being preferably
compressed before being sent to another computer (not shown
in the Figures). The images received by the second computer
will be decompressed if need be and displayed on a monitor
screen 33. In order to see a 3D stereo image, one of the two
images need to be seen by the left eye and the other image
seen by the right eye. This can be accomplished by
synchronizing the display of the two images using LCD shutter
eyeglasses (not shown in the Figures).
For motion capture function, the disparity observed by
the cameras 8, 10 is calibrated in a common 3D space of these
cameras 8, 10. The 3D position of a corresponding point
captured by two cameras can then be determined. A method for
16
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
real time disparity measurement is developed to ensure a
motion capture of limited points at quasi-video rate.
Referring to Figure 3, an object in motion can be
tracked. The rotational table 32 controlled by the computer 2
is then used in order to make the digitizer 1 follow the
object in motion. The apparatus has a limited measurement
space. In order to keep the object in the center of the
measurement space, images are acquired by the cameras 8, 10
and transferred to the computer 2. The difference between
subsequent video images is determined and used to calculate
the displacement of the object. The computer 2 controls the
rotational table 32 in order to keep the object in the center
of the measurement space.
In the motion capture mode, the computer 2 may operate
so that sequences of images in the video signals from the
first and second cameras 8, 10 are acquired in parallel.
Control points in a first image of each sequence are then
detected and tracked in the sequences of images while
disparities between the control points in the images from the
camera 8 and the images from the camera 10 are determined. 3D
positions of the control points in corresponding ones of the
images from the cameras 8, 10 taken at a same time are also
determined by using the disparities and calibration
information involving relative position and angular alignment
of the cameras. Trajectories of the control points can be
generated as sequences of the 3D positions of the control
points respectively. The control points may be formed for
example of specific markers positioned on the object or
points in the images of the object with high derivatives in
intensity.
The computer 2 may have compressing and decompressing
functions applicable on the digitized data, and functions for
transmitting the digitized data through a communication port
of the computer 2.
17
CA 02405048 2002-10-03
WO 01/81859 PCT/CA01/00560
Although white light is preferred for the projectors 3,
6, other colors may possibly be used depending on the needs.
Aperture and shutter mechanisms (not illustrated in the
Figures) can be used with the projectors 3, 6. The cameras 8,
10 used in the digitizer 1 may have adjustable gain features
if desired.
While embodiments of this invention have been
illustrated in the accompanying drawings and described above,
it will be evident to those skilled in the art that changes
and modifications may be made therein without departing from
the essence of this invention. All such modifications or
variations are believed to be within the scope of the
invention as defined by the claims appended hereto.
18
