Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
CA 02494012 2010-01-22
71073-163
WOOD TRACKING BY IDENTIFICATION OF SURFACE CHARACTERISTICS
Technical Field
[0003] The present invention relates generally to lumber or board tracking
and,
more particularly, to using unique surface characteristics ("woodprints") for
identifying individual pieces of wood and tracking them through an automated
production process with real-time information specific to the pieces or wood
such
that when a board or other piece of wood reaches a machining operation or
decision
point, its predetermined characteristic information, such as processing
information,
value, or destination, can be automatically retrieved to influence how such
piece of
wood is handled.
Background of the Invention
[0004] Conventional board tracking devices rely on spraying or imprinting
an
identification code or symbol on each board and reading the information with a
sensor after the board has traveled to a subsequent machining station. The
printing
and reading processes are preferably performed at high speeds and may be
1
,
- - CA 02494012 2005-01-24
physically difficult to reliably achieve because of the dynamic nature of the
boards
themselves.
[0005] If wood is missing from an area where the board is printed, the
board's
identification code can be illegible to the reader. A twisted board, or one
with bark or
some other defect in the print zone, can also be difficult to reliably mark
and identify.
A board may also turn over during travel between stations, requiring that
either both
sides be marked, or both sides be read.
[0006] Another problem with conventional board tracking devices is that
printing
systems contain print media, such as ink or paint, and moving parts that
contribute to
decreasing reliability. Ink jet and/or spray systems require constant
maintenance to
keep them working properly. Most require compressed air, are adversely
affected by
temperature extremes, and are very sensitive to variations in the ink or paint
quality.
For example, unless the print media is continually circulated when the system
is not
in use, the print media can freeze, its pigments can separate, and print
nozzles can
become plugged. The required maintenance can cost thousands of dollars
annually
beyond the cost of replacement parts and the original equipment itself. When a
marking system fails, the failure is typically not detected until the boards
reach the
next machine center, potentially meaning that a hundred or more boards must be
physically removed from the process and either reintroduced ahead of the
marking
system or manually processed.
[0007] Marking boards with ink or paint can reduce their value as potential
appearance-grade products destined for exposed applications. Boards typically
processed through a marking system have been previously planed and are ready
for
immediate use. Some uses include wall paneling or exposed ceilings and floors.
If
the final finish will be a non-opaque stain or paint, any non-natural marking
will not
be acceptable. Furthermore, print media typically contain or are mixed with a
fluorescent pigment or dye to provide better contrast to improve mark-reading
performance. Such pigments may be invisible in normal lighting, but the marks
will
glow under an ultraviolet (black light) source. The ultraviolet marks are,
therefore,
unacceptable for applications where the surfaces are inadvertently illuminated
by a
lighting source that emits UV light.
[0008] While existing board tracking systems may be suitable for some
specific
purposes, a more universal method for tracking boards, regardless of their
condition
or their final application, is desirable.
2
CA 02494012 2010-01-22
'71073-163
Summary of the Invention
[0009] An object of some embodiments of the present invention is,
therefore, to provide an improved wood tracking system.
[0010] Another object of some embodiments of the invention is to
provide a
surface "woodprinting TM" system and/or method for tracking pieces of wood.
[0011] An alternative object of some embodiments of the invention is
to
employ such woodprinting TM capabilities to identify individual pieces of wood
using
their unique inherent surface characteristics in order to track the wood in
real-time
through an automated production process.
[0012] Another alternative object of some embodiments of the invention is
to provide woodprinting TM capabilities that identify each piece of wood by
its
unique grain characteristics.
[0013] A further alternative object of some embodiments of the
invention is
to provide woodprinting TM capabilities that do not mark or modify the wood
surface
or its appearance.
[0014] Yet another alternative object of some embodiments of the
invention
is to provide woodprinting TM capabilities that can be used on any face, edge,
and/or end of surfaced or unsurfaced wood of any moisture content.
[0015] Still another alternative object of some embodiments of the
invention
is to provide woodprinting TM capabilities that are relatively insensitive to
lighting or
positioning differences during image data collection.
[0016] As with human fingerprints and snowflakes, wood cells in trees
develop in distinct manners as a result of many factors, including genetics,
environment, weather, soil, local life form effects, and many other
contributing
elements. The surface of an individual board or piece of wood can also be cut
from a tree in an infinite number of angles. Accordingly, the resulting
exposed
grain structure of any given surface will be unique when compared to that of
any
3
CA 02494012 2010-01-22
' '71073-163
other surface of any other piece of wood. Even boards taken from the same tree
and cut at similar angles will have unique grain structures. Furthermore, if
the
piece of wood is twisted, has some bark, is missing wood, has other physical
shape defects such as wane defects, or has other defects such as knots or
pitch,
within an area of interest, these features become additional characteristics
that
can be used to uniquely identify the piece of wood.
[0017] Some embodiments provide a surface "BoardprintTM" or
"WoodprintTM" identification technique for identifying individual pieces of
wood
using their inherent unique surface characteristics to track them in real time
through an automated
=
3a
- CA 02494012 2005-01-24
production process. Such embodiments overcome many of the disadvantages of
conventional board tracking devices. The improved tracking technique permits
information specific to a piece of wood, such as proposed machining
information,
value, destination, and/or other characteristics, to follow the individual
piece of wood
through an automated process. Thus, when the piece of wood reaches a point in
a
production process where a decision is desired, the unique identity of the
wood piece
can be used to retrieve the appropriate information already associated with
it.
[0018] Some embodiments employ cameras, lighting, image acquisition
hardware, a computer, and image processing software. Preferred processing
algorithms reduce effects from random or varying angles of image collection
and/or
from fluctuations of lighting sources. Statistical parameters, conceptually
similar to
those used for human fingerprint matching technology, provide a certain amount
of
flexibility in the analytical process. So, as long as a piece of wood meets
desired
statistical requirements for a match, the wood piece is considered to be a
match. An
adjustable tolerance for accuracy can be employed to compensate for
fluctuations in
the readability of wood grain characteristics, for example. These techniques
are
unlike conventional wood tracking techniques that attempt to find an exact
match for
a printed code.
[0019] Preferred image processing techniques of the invention can be used
on
any face, edge, and/or end of surfaced or unsurfaced wood of any moisture
content,
such as green (uncured) or dry wood, and the image processing techniques are
relatively insensitive to defects on the surfaces. So, unlike convention
tracking
systems which cannot track relatively green or unsurfaced boards that do not
facilitate the use of print media, embodiments of a woodprintTM identification
system
can be used anywhere in a sawmill and/or planer mill process where a wood
surface
can be imaged and is not limited to use with dry, surfaced boards.
Furthermore,
since printing or stamping the boards can be eliminated and the image
acquisition
sensors do not need to contact the wood surfaces, the wood is left with no
additional
markings that could degrade its appearance or adversely affect its value or
merchantability.
[0020] Preferred embodiments of the invention can be employed to work in
conjunction with an automated wood grading system. Because the scanning and
computer processing associated with automated board or lumber grading may take
several seconds, boards may travel away from the scanner and be mixed in with
4
CA 02494012 2014-08-28
71073-163
subsequent or previous boards without adversely affecting throughput. The
woodprintTM identification techniques therefore facilitate automated matching
of grade
solutions with correct boards or other pieces of wood when they are presented
later
for further machining.
[0021] Some preferred embodiments can be implemented without adding
appreciably to the production cost of wood products. Some preferred
embodiments
employ no moving parts and utilize components that are easily protected from
temperature fluctuation and other environmental concerns. Some preferred
embodiments can also utilize existing scanning hardware for initial image
acquisition
in existing systems, merely adding a software element to the scanning system.
Off-
the-shelf lighting, cameras, and/or other sensing hardware can be added
downstream
at various points in the automated production system. The expected
maintenance,
such as keeping the camera viewing ports free from debris, and occasionally
replacing of light sources, would also be minimal. Accordingly, the preferred
surface
woodprinting TM techniques for tracking wood pieces is highly reliable.
[0021a] An aspect of the invention relates to a method for tracking
multiple
pieces of wood, each piece of wood having multiple surfaces and at least an
end
surface, while inhibiting misidentification of wood having received an
alteration to its
end surface, the method comprising: obtaining first inherent wood
characteristic
information from at least a first portion of at least a first surface of each
of the multiple
pieces of wood as they move through a first station, the multiple pieces of
wood
including a first piece of wood, and the first portion including a first area
that is
located away from the end surface that is positioned to be altered in its
appearance
at a time after the first inherent wood characteristic information is
obtained;
converting at least some of the first inherent wood characteristic information
concerning the first area for each of the multiple pieces of wood into
respective first
data packets; obtaining, based on a location of the end surface, second
inherent
wood characteristic information from at least the first area of each of the
multiple
pieces of wood as they move through a second station so that the second
inherent
wood characteristic information is free from effects of alterations;
converting at least
5
CA 02494012 2014-08-28
71073-163
some of the second inherent wood characteristic information for the multiple
pieces of
wood into respective second data packets; and comparing a second data packet
to a
first data packet of the first piece of wood to make a first evaluation of
similarity
between the first and second data packets to determine whether a given piece
of
wood that has traveled through the second station is the first piece of wood.
[0021 b] Another aspect of the invention relates to a method for
tracking multiple
pieces of wood, each piece of wood having multiple surfaces, by providing
unconstricted transport of wood pieces undergoing surface characterization,
the
method comprising: moving multiple pieces of wood in a first direction of
travel
through a first station in a manner that permits the multiple pieces of wood
to move in
a direction transverse to the first direction of travel; obtaining first
inherent wood
characteristic information from at least a first portion of at least a first
surface of each
of the multiple pieces of wood as they move through the first station while
automatically compensating for movement in the direction transverse to the
first
direction of travel, the multiple pieces of wood including a first piece of
wood;
converting at least some of the first inherent wood characteristic information
concerning the first portion for each of the multiple pieces of wood into
respective first
data packets; obtaining second inherent wood characteristic information from
at least
the first portion of each of the multiple pieces of wood, including the first
piece of
wood, as they move through a second station; converting at least some of the
second
inherent wood characteristic information for the multiple pieces of wood into
respective second data packets; and comparing a second data packet to a first
data
packet of the first piece of wood to make a first evaluation of similarity
between the
first and second data packets to determine whether a given piece of wood that
has
traveled through the second station is the first piece of wood.
[0022] Additional objects and advantages of this invention will be
apparent
from the following detailed description of preferred embodiments, which
proceeds
with reference to the accompanying drawings.
5a
CA 02494012 2014-08-28
. 71073-163
Brief Description of the Drawings
[0023] FIG. 1 is a pictorial representation of exemplary desirable
components
of one embodiment of a wood tracking system.
[0024] FIG. 2 is a flow diagram of an overview of exemplary wood
tracking
events that occur in one embodiment of a wood tracking system.
[0025] FIG. 3 is a flow diagram detailing exemplary image analysis
executed at
an exemplary characterization station in one embodiment of a wood tracking
system.
[0026] FIG. 4 is a flow diagram detailing exemplary image analysis
executed at
an exemplary identification station in one embodiment of a wood tracking
system.
Detailed Description of Preferred Embodiments
[0027] FIG. 1 shows an exemplary wood tracking system 10 for
characterizing
and identifying pieces of wood 8 in a production line as they move in a
direction of
travel 6 through an automated production process to downstream processing
centers
and/or sorting bins. FIGS. 2-4 show general and specific flow diagrams of wood
5b
- - CA 02494012 2005-01-24
tracking events and analysis employed in some embodiments of wood tracking
system 10.
[0028] With reference to FIG. 1, a characterization station 12 captures and
processes an image of at least a portion of each piece of wood 8 and then
coordinates the tracking of the specific pieces of wood 8 with the automated
processing system (APS) as they travel to common or different machining
centers for
the sawing, grading, and/or subsequent sorting. The pieces of wood 8 can flip
over
during travel or arrive at the machining centers out of sequence. A downstream
identification station 42 at one or more of such machining centers eventually
captures and processes the images of most, if not all, the pieces of wood 8.
Then
image data obtained at the identification station 42 is compared to image data
obtained at the station 12 to identify each piece of wood 8 and report its
location to
the automated processing system. Each piece of wood 8 is then processed or
sorted according to wood piece-specific instructions from the automated
process
system. After being processed, each piece of wood 8 may be re-characterized by
the original characterization station 12 or another characterization station
12 that is
downstream of such processing or machining center.
[0029] General embodiments of a wood tracking system 10 include one or more
wood characterization stations 12 having one or more light sources 14a and 14b
(generically light sources 14) and corresponding image acquisition hardware
16a
and 16b (generically image acquisition hardware 16) that are directly or
indirectly in
communication with camera interface hardware 18 that is in turn directly or
indirectly
in communication with a computer 20. A typical wood characterization station
12
also preferably employs image processing software 22 having customized image
processing algorithms designed to analyze video, frame, or other captured
image
information using various data manipulation techniques described later in
detail.
[0030] Enough unique inherent information can be extracted from a small
surface
portion of each piece of wood 8 as it travels at a high speed through the
characterization station 12 to successfully characterize each piece of wood 8
uniquely. Typically, pieces of wood 8 travel (with their major axis in the
direction of
travel 6 (lineally)) through station 12 at speeds of about 91 to 915 meters
(300 to
3000 feet) per minute and more typically at speeds of at least 366 meters
(1200 feet)
per minute and less than about 610 meters (2000 feet) per minute. These speeds
can be increased when the pieces of wood 8 are more homogeneous and when
6
- - CA 02494012 2005-01-24
material handling challenges are solved. In some embodiments, a roughly 61-
centimeter (two-foot) surface portion that is about one meter (three feet)
from the
end of each piece of wood 8 is imaged at station 12. In other embodiments, two
or
more portions are collected from one or more surfaces of each piece of wood 8
that
is imaged at station 12.
[0031] For convenience, the terms "wood," "pieces of wood," or "wood
pieces"
may refer to pieces of timber, lumber, logs, flitches, cants, beams, posts,
studs,
boards, veneer, and/or any other pieces of wood smaller than the whole tree
and
larger than sawdust. The surfaces characterized may be one or more faces 30,
sides 32, and/or ends 34 of the wood pieces 8. Skilled persons will appreciate
that
curved or intersecting surfaces may additionally or alternatively be
characterized at
station 12. Skilled persons will appreciate that the minimum area that is
imaged
should be an amount of area sufficient to yield a unique recognizable vector
file or
image packet as later described. Skilled persons will further appreciate that
the
minimum area imaged may be different for different wood products or may be
different for wood pieces having different shapes or sizes. Alternatively, any
additional amount of any wood surface can be imaged, including up to all
portions of
all surfaces. Skilled persons will also note that any combination of surfaces
may
also be imaged.
[0032] In some embodiments, the image captured comprises a high contrast
image of grain and/or growth ring characteristics of a portion of each piece
of wood
8. These features are relatively insensitive to differences in lighting,
position, and
contrast, but other wood characteristics such as tracheid patterns and/or wood
defects could be used for identification purposes. Such defects might include,
but
are not limited to, physical shape defects such as cups, crooks, cracks,
knots, wane,
twists, bark, and/or pitch.
[0033] In some embodiments, images or other inherent wood characteristic
information is obtained from at least two or more discrete portions of each
piece of
wood 8 at station 12. This additional information could be useful when a piece
of
wood 8 is cut into shorter pieces, some or all of which could then be
identified
downstream. This additional information could also be used as additional data
set
for a given piece of wood 8 to increase identification accuracy downstream.
[0034] Although for some embodiments only one or more portions of one surface
of each piece of wood 8 may be used for analysis at the characterization
station 12,
7
- - CA 02494012 2005-01-24
images of one or more corresponding portions of opposite surfaces can also be
collected at the characterization station 12 so that opposite surface
information is
available in the event that a piece of wood 8 is turned over in transit
between the
characterization station 12 and the identification station 42. Accordingly,
the image
acquisition hardware 16 and the respective lighting sources 14 can be mounted
in
such a way as to obtain images of portions of opposing surfaces of each piece
of
wood 8 as it travels through station 12.
[0035] Appropriate gaps may be provided between conveyor belts or other wood
conveying means to permit viewing of the bottom surface of each piece of wood
8.
In alternative embodiments, the sets of image acquisition hardware 16 and
light
sources 14 may be symmetrically positioned about opposite surfaces of wood
pieces
8 to characterize opposite surfaces simultaneously or may be positioned at a
distance so that the opposite surfaces are characterized sequentially.
Alternatively,
pieces of wood 8 may intentionally be flipped over and re-sent through a
station 12
having only a single set of lighting sources 14 and image acquisition hardware
16 to
characterize an opposite surface, or the image acquisition hardware 16 could
be
moved such that its field of view changed to include the opposite surface.
However,
such embodiments might entail additional safeguards to ensure that images of
the
opposite surfaces are properly correlated with each other. The images or image
data collected at station 12 from some or all of the surfaces may be
communicated
to the same computer 20 for processing.
[0036] In one embodiment, process blocks 110a and 110b employ image
acquisition hardware 16 that comprises one or more monochrome array cameras
with at least about a 1024 x 768 pixel addressability. Such camera(s) can be
configured to be electronically shuttered up to at about at least 7.5
frames/second
with an adjustable integration time to include 500 microseconds, or such
camera(s)
can be configured to work with a strobed light source 14 having similar strobe
frequencies. In an exemplary embodiment, a camera is positioned a minimum of
61
centimeters (24 inches) over a face 30 of a target piece of wood 8 to provide
a
minimum field of view of 36 centimeters by 61 centimeters (14 inches by 24
inches)
using a fixed lens. Skilled persons will appreciate that the camera speed and
resolution can vary depending on the speed of the production line and the
maximum
field of view desired.
8
CA 02494012 2010-01-22
^71073-163
[0037] In some circumstances, the area toward the end of a piece of wood 8
may
not be ideally suited for identification analysis because it may be used by
operators
to add additional information using crayon marks, for example. If so, the
crayon
marks could possibly reduce the success of identification analysis as such
marks
could mask significant surface characteristics on the piece of wood 8 and/or
this area
may be cut before the piece of wood 8 arrives at station 42. Therefore, even
though
an image or characteristic information can be obtained from any location along
the
surfaces of the piece of wood 8, some embodiments are configured at both
stations
12 and 42 to use an area at least 30 cm (12 inches) away from the end of the
piece
of wood 8, more preferably at least 61 cm (24 inches), and most preferably at
least
760 cm (30 inches).
[0038] In some embodiments, however, it can be more advantageous to take
images at the station 42 that include the end of the board. The video
processing
software can then reposition the image with respect to the end of the board.
In
particular embodiments, the camera and lighting are configured to take an
image of
the top face of the board at the even-ended end (the lumber line end) and
include in
the image approximately one inch past the end of the board. The image is then
analyzed to not only find the edges of the board, but to also find the end of
the
board, and then the part of the image that is actually used for image data
analysis
starts at a specific distance from the end of the board, such as ten
centimeters (four
inches), to allow for aberrations in the even-ending process.
[0039] For most embodiments, it is desirable that the positions of images
taken at
both of the stations 12 and 42 match with each other since most techniques
used for
matching work better when the relative locations of the characteristics are
known.
Accordingly, it is desirable for the position of the piece of wood to be
accurately
located at both stations 12 and 42.
[0040] In some embodiments, the piece of wood 8 may be traveling lineally
at
station.12, so an encoder can be used to track the position of the piece of
wood 8
and reference the image(s) taken. The pieces of wood 8 are preferably allowed
to
"float" through station 12 such that they can move sideways (transverse) a
several
inches, but create no image processing difficulties since such movement is
accommodated in a normalization process described later. At station 42, the
pieces
of wood 6 may be traveling in the transverse direction, and the piece of wood
8 can
be brought against a fence using even-ending rolls. Their position in the
width
9
CA 02494012 2010-01-22
71073-163
direction can also be tracked with an encoder, and although the normalization
process described later can account for some movement in this direction, the
lineal
position of the board is more preferred, so even-ending rolls and a fence are
employed.
[0041] Selective use of one or more color channels or one or more filters
(not
shown) may reduce or eliminate infra-red energy and all visible wavelengths
except
green because most wood features of interest provide better visible contrast
in green
wavelengths. One or more filters may also be added to reduce or eliminate
other
unwanted characteristics of ambient lighting to enhance desirable effects of
the light
source(s) 14. The image acquisition hardware 16, and/or any other sensitive
components as appropriate, may be housed in a NEMA 12/13-rated enclosure to
protect it from temperature extremes, dust, and dirt. All of the components
described
above are commercially available.
[0042] In an alternative embodiment in which the wood piece 8 is a board, a
SonyTM XCD-X700 monochrome camera or a color camera is mounted about 1.5
meters (five feet) above the face 30 of the board and is preferably oriented
substantially perpendicularly to the board. The camera is installed in a NEMA
12/13-
rated enclosure. The camera has a Fire Wire serial communication port, and a
Fire
Wire extender converts the Fire Wire signal to propagate through a fiber optic
medium over a longer distance than the camera's own circuitry permits.
Filtering
can be implemented to allow only green energy to reach the camera or selective
channels can be used if wavelength discrimination is desirable, and the
background
behind the board is preferably dark to facilitate image collection. In some
embodiments, the image acquisition cycle is triggered by a photoeye proximity
transition sensor that detects the end 34 or edge 32 of the board, depending
on the
orientation of the board as it travels through the station 12. Software
running in the
camera automatically transmits the acquired data to a personal computer 20.
[0043] In another embodiment, one or more color high-speed line scan
cameras,
having separate red, blue, and green channels, capture an image from a surface
as
a piece of wood 8 moves through station 12. The image data from the green
channel is pieced back together to create an array image for subsequent
analysis
with software tools. In another embodiment, a single 768 x 768 pixel
monochrome
camera is employed to capture an image of an end 34 of a piece of wood 8 in a
field
of view of approximately 36 centimeters by 36 centimeters (14 inches by 14
inches)
- - CA 02494012 2005-01-24
at the first station 12. It is desirable to capture a high quality image, but
the
particular method of image data capture employed may not be particularly
relevant.
Skilled persons will appreciate that successful image collection is not
limited to using
imaging techniques described above and that numerous types of image
acquisition
hardware are commercially available and numerous configurations and positions
could be employed. For example, other color or black and white image data may
be
obtained, or even radio or other frequency data may be obtained and used for
"image matching." In addition, since image data can be available from any
location
along any surfaces of the piece of wood 8 as described in this embodiment, any
or
all portions of the board surface can be used for image matching.
[0044] Skilled persons will appreciate that image acquisition hardware may
take
on many other forms. In one embodiment, photodiodes are employed to capture
inherent characteristic information obtained from exposing the wood pieces 8
to x-
ray radiation. In another embodiment, antennas are employed to capture
inherent
characteristic information obtained from exposing the pieces of wood 8 to
radio
waves. Skilled persons will also appreciate that that more than one imaging
technique can be used at the same time, combining multiple sensors and/or
energy
sources to ultimately improve the success of the image data-matching task.
[0045] Light sources 14 may employ almost any commercially available
lighting
equipment or known lighting technique. Desirable light sources 14 provide
wavelengths generally considered to be green light at sufficient energy to
accommodate the integration time necessary to capture well-contrasted images.
Too much light is not likely to be a problem since the camera aperture or
integration
time can be adjusted to compensate for excess intensity. Although light
sources 14
with reflector and lens systems that project a visible intensity pattern onto
a wood
surface can be employed, lighting embodiments that do not project visible
intensity
pattern onto a wood surface are generally easier to implement. In one
embodiment,
station 12 employs one or more a broad-spectrum lamps, such as commercially
available halogen or Xenon lamps that consume about 1000 Watts of direct
current
(DC) power to generate approximately 800 Watts of light energy. The lamp's
self-
contained lens and reflector cast a diffuse light over an approximate area of
about 76
centimeters by 76 centimeters (30 inches by 30 inches). The visible energy
produced by such a light source 14 is "white," including energy in green
wavelengths
that is in proportion to the energy in other visible wavelengths.
11
- - CA 02494012 2005-01-24
[0046] In another embodiment where the wood piece 8 is a board, two 1000
Watt
broad spectrum stage lights are mounted about 1.8 meters (six feet) away from
the
face 30 of the board, generally in a longitudinal plane of the board that is
perpendicular to the face 30 of the board and generally at 40 degree angle off
the
face 30, to reduce reflections and provide a fairly uniform illumination over
a 0.6
meters square (two foot square) area of the board face 30. Such lighting and
camera combination can be duplicated as necessary in order to capture more
sections of a board's surface for analysis.
[0047] Skilled persons will appreciate that different lighting techniques
can be
employed to accommodate different image acquisition hardware 16. If a line
scan
camera is employed, the light source 14 should be bright enough in the green
portion
of the visible spectrum, for example, to provide for successful image
integration. In
another example, a strobe light source 14 is employed to control the camera's
acquisition of the image instead of depending entirely on a camera's
electronic
shutter mode. Such light source should also generate sufficient light in the
green
wavelengths of the visible spectrum to acquire a successful image for
processing.
Skilled persons will appreciate that green laser or other coherent light
sources 14
can be employed and may be preferred when the costs of such embodiments
become comparable to the costs of more typical light sources such as those
discussed above.
[0048] Process blocks 112a and 112b may employ camera interface hardware 18
that preferably resides in a computer 20 and may be a circuit board or card
that
captures the resulting image from the image acquisition hardware 16 for
analysis in
the computer 20 and that may also control activation of image data acquisition
and/or strobe light timing. Such a circuit board is often referred to as a
"frame
grabber," is commercially available, and may be located in the computer's
system
bus. Preferred functions of such card include the capability to properly
exercise real-
time imaging functions of the image acquisition hardware 16, capture the
resulting
image data, and present the data to the computer 20 for analysis. For some
embodiments, such a card is preferably capable of receiving an external signal
to
initiate each image capture, and then triggering the strobe and camera to
initiate
image capture.
[0049] Skilled persons will appreciate that if a different kind of camera
is used, the
camera interface hardware 18 will be changed appropriately. For example, if
the
12
- - CA 02494012 2005-01-24
image acquisition hardware 16 were a line scan device, then the camera
interface
hardware 18 would collect image data in single scanned lines and present the
data
to the computer 20 for assembly into an array image for evaluation. Skilled
persons
will also appreciate that such card may facilitate only the data collection
process and
not control the camera. Another example can use a camera containing its own
acquisition software and hardware, whereby the data is subsequently
transmitted to
the computer 20 via a serial communication link such as Ethernet, USB, or Fire
Wire,
requiring no special hardware to be installed in the computer 20.
[0050] Skilled persons will appreciate that camera interface hardware 18
may
reside in image acquisition hardware 16 or in the personal computer or may be
an
independent device that is directly or indirectly interfaced to the image
acquisition
hardware 16 and/or the computer 20. As cameras and computers 20 are continuing
to evolve and take on a more complete roles to ease integration, many or all
of the
functions camera interface hardware 18 will be incorporated into the image
acquisition hardware 16, such that a high-speed communication port in the
computer's motherboard will accept the data directly from the image
acquisition
hardware 16 so separate camera interface hardware 18 may be omitted entirely.
[0051] In some embodiments, computer 20 may be an off-the-shelf personal
computer (PC) with commercially available hardware and software components.
Such a computer 20 may have typical attributes, such as a processor speed of 2
to 3
gigahertz, minimum RAM memory 512 megabytes, hard disk, CD drive, keyboard
(kbd), mouse, color monitor, Ethernet network card, Microsoft Windows
operating
system, an interface hardware card, and commercially available and/or
customized
image processing software and other application(s).
[0052] In accordance with process block 114a, the computer 20 processes the
image data collected from the image acquisition hardware 16 associated with
station
12, and in accordance with process blocks 116 and 118 then communicates the
results via an Ethernet connection to other stations 42, production line
equipment, an
automated processing system (APS), and/or possibly other computers. The
connections between the various components and stations can take on many
forms,
the most common being coaxial cable or shielded twisted paired cables. In
addition,
the communication medium could be via fiber optic (light energy) or radio
frequency.
The types of connections used may depend on the proposed communication method
13
- - CA 02494012 2005-01-24
desired and the ability to integrate with preexisting components of older wood
processing systems, and numerous specific embodiments could be implemented.
[0053] The image or image data can be converted to an "image packet"
(described later) to circumvent current technological limitations, such cost
constraints
associated with communication bandwidth and computational power, that hinder
the
transmission of complete wood piece images and the use of direct image to
image
comparisons. In some embodiments, the image packet, and not the raw image
data,
is transmitted to the computer 20 to reduce communication time, and the image
packet is what is compared, not the raw or compressed image. The image
conversion and comparison preferably operates in "real-time," at least as fast
as the
production line to which the wood tracking system and method are connected.
[0054] The computer hardware employed is expected to be changed over time to
incorporate higher speed components and interface devices, as well as
increased
memory capacity when appropriate as improved components become available. For
example, an Ethernet 100/10 Base T network can be replaced by a Gigabit
Network
or other serial or parallel networks. These communication interfaces can also
be
integrated into the computer's motherboard, eliminating the need for adding a
separate card for the communication function. Fiber optic or radio frequency
mediums are also potential methods for this function. The computer 20 is
connected
to a power source that is preferably protected from voltage or current
fluctuations.
The image acquisition hardware 16 and light sources 14 may be connected to the
same power source as the computer 20 or may be connected to separate power
sources.
[0055] With the same or different sensors and image processing equipment, the
automated processing system examines each piece of wood 8 to determine the
most
beneficial use of each piece of wood 8, such as grade and dimension, and to
provide
the corresponding processing instructions, such as where to cut. Such
examination
may be accomplished concurrently with characterization for identification, or
such
examination may occur before or after ID characterization or at a separate
location.
Conventional automated processing systems may employ some or all of the
techniques disclosed in any one of U.S. Patent Nos. 6,757,058; 6,756,789;
5,703,960; 5,644,392; 5,524,771; 5,412,220; 5,254,859; 5,252,836; 4,992,949;
4,926,350; 4,916,629; 4,879,752; 4,867,213; 4,831,545; 4,827,142; 4,606, 654;
14
_..
_
CA 02494012 2013-09-25
71073-163
4,301,373; 4,286,880; 4,221,974; 4,207,472; 4,086,496; or combinations
thereof.
[0056] With reference again to FIGS. 1-4, an identification station 42
that is later
or downstream of the characterization station 12 may employ types and
embodiments of light source(s) 14, image acquisition hardware 16, camera
interface
hardware 18, and/or computer(s) 20 that are identical to, or different from,
those
employed in the previous station 12. Skilled persons will appreciate that in
most
cases identical types and embodiments of these components are preferred to
facilitate ease in matching the image data of wood surface characteristics
obtained
at stations 12 and 42.
[0057] An image may be captured of an entire surface or from only a specific
area
on the piece of wood 8 at both stations 12 and 42, and/or identification
station 42
may capture fewer and/or smaller images of portions of wood pieces 8 than
those
captured at the characterization station 12. Skilled persons will appreciate
that if
station 12 obtains image data from both of the opposite surfaces of each piece
of
wood 8, then station 42 will preferably obtain image data from only one of the
opposite surfaces in order to establish a match. In some embodiments, pieces
of
wood travel through station 42 in an intentionally different orientation than
the
orientation in which they travel through station 12. For example, instead of
lineal
travel through stations 12 and 42, pieces of wood 8 may have their major axis
oriented transversely to, and preferably perpendicularly to, direction of
travel 6 at one
or both stations 12 and 42. Skilled persons will also appreciate that
transverse travel
speeds may be different from lineal travel speeds. Thus, the processing of the
images from stations 12 and 42 preferably take into account speed differences
as
well as different image acquisition techniques in order to successfully
support a
BoardprintTM matching schema.
[0058] Preferred embodiments employ a computer 20 at each downstream
identification station 42. Each computer 20 is preferably capable of
performing the
same tasks concerning collecting image data of wood surface characteristics
from its
respective image acquisition hardware, with either computer 20 being able to
make
the final comparison decision. Alternatively, downstream computers 20 may be
adapted to perform only some or all of the image characterization, but leave
the
comparison processing to a central computer (not shown) or a computer 20 at a
- - CA 02494012 2005-01-24
previous station. The computers 20 preferably communicate with each other and
the
automated processing system via an Ethernet or other high-speed communication
network such as those described above.
[0059] Skilled person will appreciate that a single high-speed computer 20
may
perform the image characterization and comparison tasks associated with two or
more stations 12 and 42. Such a computer 20 would preferably contain the
camera
interface hardware for the image acquisition hardware of each station 12 and
42 and
would be configured to multitask between the demands of each. Some pre-
existing
wood processing systems already have, for example, a complete scanning system
in
place that includes its own light sources 14, cameras, and computers 20 with a
software architecture that facilitates image acquisition in one computer,
image
processing in another, decision logic in a third, and communication to the
production
equipment in another. Skilled persons will also appreciate that the image data
collection, processing, and decision process can be implemented in multiple
computers 20 that are remote from the stations 12 or 42 where the image
acquisition
hardware is located. Skilled persons will further appreciate that station 42
may be at
a remote location from station 12 such as a separate milling site or customer
plant
with data sent over the internet or other means, and station 42 may be
operated
shortly after or at a much later time from when station 12 acquires the image
data of
a wood piece 8. The physical location and configurations of the data
collection and
processing hardware can be varied greatly so long as such implementations are
reliable and facilitate and desirable processing and communication tasks and
speeds.
[0060] When a computer 20 receives an image from the image acquisition
hardware 16, the image and/or data associated with it may be stored in a
circular
buffer by software running in the computer 20 at least until the image data is
received and/or acknowledged by the respective decisions block 120a or 120b of
the
image processing hardware. Customized image processing algorithms may be
employed to analyze image data with various data manipulation techniques. Such
software is preferably adapted to operate in the WindowsTM environment and
integrate with communication applications in order to share data and results
with
other computers 20. The software is preferably adapted to take advantage of
Intel's
Image Processing Library (IPL), Integrated Performance Primitives (IPP),
and/or
other available libraries that feature specialized software routines that take
16
- CA 02494012 2005-01-24
advantage of specific processing hardware of the computer 20. One advantage of
using these libraries is speed. A function call to an IPL routine typically
runs much
faster than improvised dedicated C++ code. In one embodiment where the wood
pieces 8 are boards, the image processing software includes C++ libraries
and/or
Visual Basic OCX / DLL software provided by a company called Imaging Control
of
Charlotte, North Carolina. Another software application resident in the
computer 20
moves the image out of the buffer to a separate memory location so it can be
easily
accessed by the image processing software.
[0061] The software is preferably adapted to perform one or more of the
following
applications, more preferably two or more of these applications, and most
preferably
all of these applications: image warping; Fast Fourier Transformation and/or
other
transformation techniques such as wavelet analysis; interim solution
communication; woodprint matching; and/or final solution communication. A
description of each of these applications is described below.
[0062] With reference to process blocks 114a and 114b, image warping
addresses differences between the images taken by the image acquisition
hardware
16 at each station 12 or 42. Such differences include, but are not limited to:
the
position of where each piece of wood 8 ends up in the field of view;
resolution
differences between the individual image acquisition hardware 16 at different
stations 12 and 42; and image contrast. The software will take the image
received
from the image acquisition hardware 16 and scale it to correspond to a pre-
determined pixel/inch value in accordance with process blocks 124a and 124b.
The
software will also adjust the orientation of the piece of wood's image to
compensate
for any skewing or offset in accordance with process blocks 122a and 122b. The
scaling and orientation adjustments can be done in either order. A calibration
procedure at installation using a target with known size and orientation
defines the
reference points and adjusts for differences in focus, pixel size, lighting,
and skew
using a calibration target with known size, orientation, and/or other
features.
[0063] In an exemplary system, where the images taken at stations 12 and 42
are
adapted to be appear as similar as possible, the focus, illumination, and
spatial
alignment are precisely adjusted. Such calibration can be entirely manual or
can be
partially automated. In an exemplary calibration process for such a system
where
the piece of wood 8 is a board, after the camera and light sources are
mounted, a
calibration target is placed under the camera. This target may be a board with
lines
17
- - CA 02494012 2005-01-24
drawn on it using a green or black marker that border a region of interest.
The
camera is placed in a mode that makes it continually acquire images. The
focus,
aperture, and speed (acquisition time) are adjusted to produce a clear image.
The
position of the camera is then manually adjusted at both stations 12 and 42 so
that
the cameras produce nearly identical images. Focus is readjusted as needed.
[0064] In an exemplary image warping application that addresses scaling
wherein
the wood piece 8 is a board, the computer software thresholds multiple (or
each)
horizontal video lines across the image to find the edges of sides 32 of the
board. In
some embodiments, the image analysis techniques used to find one or more of
the
edges 32 of the image of boards automatically compensate for side-to-side
movement while the boards travel through station 12. Accordingly, the boards
do not
have to be tightly controlled as they move lineally through station 12, such
as by
using a guide system such as an anvil or constricting fence along a second
side.
The same imaging techniques can be used for the station 42, and thereby also
reduce the dependence of the system on knowing the lateral position of the
board.
[0065] The data from the longest lines (representing the widest part of the
board)
are used in a least-squares fit calculation to create an artificial border
that represents
the edges of the board. Using the longest lines helps eliminate areas of
missing
wood and dark fiber that can make the board appear narrow. The calculation
results
in a border rectangle that corresponds to the edges of the board and the
portion of
the field of view that was calibrated for both stations 12 and 42. The image
is then
scaled to a size compatible to both stations 12 and 42 using a bi-linear
transformation. The area within the border rectangle is next divided into two
grids: a
fine grid and a coarse grid. In an example employing a 25.4-cm (ten-inch) wide
board, the fine grid comprises 100 rectangles or cells and the coarse grid
comprises
25 rectangles or cells. In some embodiments, each fine cell may, for example,
represent a 64 x 64 camera pixel area.
[0066] In an exemplary image analysis process, each cell may be represented
by
a set of values determined by a Fast Fourier Transform algorithm in accordance
with
process blocks 126a and 126b. Exemplary values of interest include the most
often
found frequency in a grain pattern or growth ring pattern, a representation of
such
frequency's energy or abundance (magnitude of predominant spatial frequency)
compared to those of any other frequencies, and an indication of the grain
direction
and/or grain direction confidence. Other inherent wood characteristics that
may be
18
CA 02494012 2013-09-25
71 073-1 63
=
evaluated include, but are not limited to, contrast, density (as obtained by
radio
wave or x-ray analysis), or wane characteristics.
[0067] Values describing two or more of these features are determined for each
cell within both the fine and course grids of the image. Two currently
preferred
values are grain direction and magnitude of its spatial frequency because
these
values tend to be fairly insensitive to light intensity and contrast. The
actual numbers
resulting from the transformation may then be thresholded dynamically to
compensate for brightness differences between stations 12 and 42, if
desirable. In
one example, both grain direction and predominant spatial frequency are
allocated
numbers between 0 and 255. The collection of the resulting numbers for some or
all
of the cells for each piece of wood 8 is called a vector file or "image
packet."
[0068] Other well-known transformation methods such as wavelet analysis may
additionally or alternatively be utilized to render the image data. Some
examples of
wavelet transforms and their use in image storage and analysis are described
in
detail in U.S. Pat. No. 5,710,835.
[0069] These techniques take advantage of grain patterns of wood,
which like
human fingerprints are unique to each individual. In accordance with process
blocks
128a and 128b, additional imaging, filtering, and analytical techniques can
also be
employed for identifying repeating or non-repeating features of each piece of
wood 8
and/or modifying image packet for the purpose of characterizing and later
identifying
a piece of wood 8. These include any characteristics captured or analyzed by
the
automated processing system or characteristics scanned and analyzed as
disclosed
in any one of U.S. Patent Nos. 6,757,058; 6,756,789; 5,703,960; 5,644,392;
5,524,771; 5,412,220; 5,254,859; 5,252,836; 4,992,949; 4,926,350; 4,916,629;
4,879,752; 4,867,213; 4,831,545; 4,827,142; 4,606, 654; 4,301,373; 4,286,880;
4,221,974; 4,207,472; 4,086,496; or combinations thereof.
[0070] In accordance with process blocks 116 and 118, the image
packets are
sent to the automated processing system and associated with a wood piece
number
and sent to the computer 20 that will later perform the identification
comparison. In
some particular embodiments, the image packets may be save in a first-in/first-
out
(FIFO) packet queue before they are associated with a wood piece number as
indicated in process block 130. The interim solution communication involves
the
destination and application(s) for the image data once it has been transformed
to an
19
CA 02494012 2010-01-22
*71073-163
image packet. Typically, an image packet's destination will be determined by
which
station 12 or 42 collected the image. If the image data is collected by a
station 12
and an initial characterization image packet is created, a sequential wood
piece
number (WPN) or code is assigned to the image packet, and the information is
sent
to a computer 20 in communication with a downstream station 42 via the
Ethernet
connection in accordance with process block 116. In accordance with process
block
118, the wood piece number is also made available to the automated processing
system so it can relate the wood piece number to its own information about the
piece
of wood 8. Alternatively, skilled persons will appreciate that the wood piece
number
could be generated by the automated processing system and communicated to
station 12.
[0071] In accordance with decision block 140, the computer 20 associated
with
station 42 waits to receive the characterization image packet and may store it
in a
buffer memory in a sequential queue of image packets as indicated in process
block
142 to associate it with a wood piece number. In a particular embodiment with
reference to FIG. 3, the software at one of the computers 20 waits to receive
a wood
piece number from the automated processing system in accordance with decision
block 150 and saves the wood piece number in a first-in/first-out queue in
accordance with process block 152. In accordance with a decision block 154,
the
software then checks whether a reset has been received from the automated
processing system or an operator keyboard. If a reset has been received, then
the
wood piece number first-in/first-out queue may be reset to zero in accordance
with
process block 156. If no wood piece number reset is indicated, then the
software
checks for an image packet reset from the automated processing system or an
operator keyboard as indicated in decision block 158. If a reset is indicated,
then the
image packet queue may be rest to zero as indicated in process block 160. If
no reset
is indicated, then the software checks to determine whether an image packet is
available in accordance with decision block 162.
[0072] If an image packet is available, the software confirms in accordance
with
decision block 164 that a valid wood piece number is available, and if so, the
software assigns the wood piece number from process block 152 to the image
packet from process block 130 in accordance with process block 166. The
software
also transmits the wood piece number to the automated process system and
increments the wood piece number queue. Finally, in accordance with process
block
- - CA 02494012 2005-01-24
168, the software transmits the wood piece number and the image packet to the
computer 20 at station 42 and increments the image packet queue.
[0073] If the image packet is collected at station 42 and a downstream or
identification image packet is created, the software proceeds with woodprint
matching in accordance with process block 144. In one embodiment, the software
at
a downstream computer checks to determine whether a wood piece number and
image packet have been received from station 12 in accordance with decision
block
170 and saves this data to a first-in/first out packet queue in accordance
with
process block 172. Then the software checks for a reset from the automated
processing system of an operator keyboard in accordance with process block
174. A
reset is performed if indicated in accordance with process block 176, or the
upstream
characterization image packet is compared to a downstream identification
packet as
described later in greater detail.
[0074] The downstream image packet is then compared to the available
characterization image packets in memory to find a match. The wood piece
number
and its location in memory allow the software to search forward and backward
in a
logical fashion to look for the image packets of pieces of wood 8 where they
are
most likely to be, and then widen so as to find those that have moved out of
order.
Initial image packets concerning pieces of wood 8 whose identities have
already
been confirmed by the downstream station 42 can be tagged accordingly or
discarded so they do not have to be compared. The matching comparison
preferably uses a weighted statistical method to compare a downstream image
packet to initial image packets in memory. Typically two or more data values
for
each cell of an image are compared to corresponding data values of cells in an
image packet in memory to look for the numbers to match within a predetermined
tolerance. If enough matches are successful for a specified number of cells, a
successful piece of wood identity is assumed. If there are not enough matches,
then
the next image packet in memory is considered. This process continues until a
match is found, a time limit is reached, or the availability of untested
initial image
packets is exhausted.
[0075] In one embodiment, the comparison process may be driven by the
expected flow of wood pieces 8. If the wood pieces 8 are expected to stay in
order
as they travel from one station 12 to one station 42, then it may be
reasonable to
only confirm that the wood piece 8 arriving at the identification station 42
is the
21
- CA 02494012 2005-01-24
expected wood piece 8. In a rudimentary embodiment, a simple accept/reject
decision might be sufficient and may be obtained by thresholding the results,
and a
reject result might be sent directly to the automated processing system in
accordance with process block 184a and/or examination or intervention by an
operator might be requested.
[0076] An exemplary image packet comparison technique described below
employs a cell-by-cell comparison scheme that awards points in accordance with
the
cell test results. The grain direction number from a specific cell derived
from the
image taken at station 12 is compared to the corresponding cell derived from
the
image taken at station 42. If the numbers match within a very tight tolerance,
such
as + 1 or up to + 5, then 3 points are awarded for the test, for example. If
they
match within a wider tolerance, such as + 6-10, then 2 points are awarded, for
example, etc. The numbers representing the magnitudes of the spatial
frequencies
or other wood piece characteristics can be similarly evaluated.
[0077] If multiple tests, such as for grain direction and frequency
magnitude, for a
cell achieve 2 or more points, for example, an additional point can be awarded
to the
cell score. Furthermore, if adjacent cell scores are high, an additional point
can be
added to the cell score. If the number or vector values of individual cells
are low
because the original image had very little grain contrast or the imaged area
was very
irregular, the scores for these cells could also be increased to increase
their
importance. The total cell scores for each image packet can then be normalized
to a
100% scale. The host computer 20 uses the normalized test cell scores to
determine if the piece of wood 8 is indeed the same piece of wood seen at both
stations 12 and 42. In one embodiment, the normalized test cell scores are
converted to an overall confidence level for the piece of wood 8 being
identified, and
such confidence level may also be affected by the confidence level of the
pieces of
wood 8 that surround the piece of wood being identified.
[0078] In more advanced embodiments, the process may be designed to
automatically recover from a problem. For example, if the identification image
packet of the wood piece 8 undergoing identification does not present a high
confidence match to the image packet of the expected wood piece 8, then the
software may check for the presence of additional image packets in accordance
with
decision block 186, and an image packet for an opposite or other surface of
the
expected wood piece 8 may be compared to the identification image packet. If a
22
- - CA 02494012 2005-01-24
. .
high confidence match is still not found, then it is possible that the piece
of wood 8
may have gotten out of position by trading places with its neighbor. Then the
software may check for the presence of additional image packets in accordance
with
decision block 186, so a comparison can be done between the expected wood
piece's image packet and its neighbor's image packet(s). These multiple-
comparison results could be then further compared using probability techniques
well
known to skilled practitioners to compensate for process particulars. For
example,
the software may look for and react to changes in confidence level patterns.
[0079] Skilled persons will appreciate that each piece of wood 8 may be
associated with two or more image packets if separate packets are desired for
each
opposite surface or for each face, side, and/or end surface or that each piece
of
wood 8 may be associated with a single image packet containing characteristic
image data for some or all of the wood's surfaces for which an image is
collected.
The additional image packets associated with each piece of wood 8 can be used
for
comparison before the image packets concerning other pieces of wood 8 are
compared to allow for possibilities of a piece of wood being turned over in
its travel
between stations12 and 42. Skilled persons will appreciate that initial and
downstream image packets may include data from one or more entire surfaces of
a
piece of wood 8 or from only portions of one or more surfaces. In some
embodiments, stations 12 convert larger portions of wood piece surface images
into
image packets and stations 42 convert smaller portions of wood piece surface
images a into image packets. Preferred minimum areas converted into image
packets should have sufficient size to reduce inaccurate identifications.
[0080] In alternative embodiments, characteristic information from
multiple
portions of one or more surfaces can be used to increase the confidence of a
correct
match and thus improve tracking success. For example, using two portions of a
board's surface for matching can be superior to using one portion alone, and
three
portions more successful than two portions, etc. Similarly, using multiple
sensing
techniques can improve tracking success. For example, using color content
along
with grain characteristics can improve tracking success, and this can be
accomplished using the same hardware. Another example is that of using a radio
frequency sensor to capture dielectric information in combination with a
monochrome
camera for grain characteristics. Some of the many possible techniques that
could
be mixed and matched include board geometry, tracheid effect from laser
radiation,
23
- CA 02494012 2005-01-24
moisture content, ultrasonic response, weight, and density patterns from x-ray
energy analysis. The addition of any information will have an overall effect
of making
the identification system more accurate.
[0081] Once a match is, or is not, found, in accordance with decision
blocks 180a
and 180b, the final solution communication involves making the information
available
to the automated process system. A successful match is signaled by presenting
the
matching sequential board number to the automated processing system via
Ethernet
connection in accordance with process blocks 182a and 182b. A non-match is
communicated by presenting a specific failure code in accordance with process
blocks 184a and 184b. Additional algorithms can be used to complement those
described. Some wood piece surfaces will be darker or offer less contrast.
Sometimes these can be enhanced using additional filtering methods. It may be
desirable to modify the image processing algorithms by adding other filtering
techniques in order to make the system more robust to accommodate such
variables. Additional sensor types and image acquisition hardware
configurations at
both stations 12 and 42 can be implemented along with modifications for
processing
the image data.
[0082] For example, if the data collected by the image acquisition hardware
16 is
presented to the computer 20 or its image collection hardware 18 in a manner
inconsistent with existing equipment and processing software, the image
processing
software can be adapted to integrate with it. As computers continue to evolve,
there
will be more options available for image processing, communication, and other
software tasks. Skilled persons will appreciate that the software will evolve
with the
computer platform hardware as the resulting architecture or speed and memory
enhancements make other software techniques favorable.
[0083] In some embodiments, all or major portions of the image processing
software reside in a single computer 20. In other embodiments, all or major
portions
of the image processing software reside in each computer 20. In yet other
embodiments, specific image processing tasks may be allocated to specific
computers 20 or spread out over multiple computers 20. In a further
embodiment,
some computers 20 may contain all or major portions of the image processing
software while other computers 20 may perform only minor tasks. Another
specific
example includes taking the image acquisition and processing tasks that would
be
done in station 12 and integrating them into a scanning system that acquires
wood
24
- - CA 02494012 2005-01-24
piece images for other purposes. The scanning system would produce the image
packet information along with its own tasks and either keep such information
for later
comparison to the downstream identification images, or pass the image packets
along to station 42 with a sequential wood piece number.
[0084] In systems with a single computer 20, the image acquisition hardware
16
from both stations 12 and 42 are connected directly or indirectly to the
computer 20
and processed with the image processing software installed on the computer 20.
This computer 20 would also be connected to the production equipment and
possibly
other computers 20. Additional cabling configurations are possible that
connect the
computer 20 with a separate image collection/processing system.
[0085] In addition, preferred embodiments employ quality control components
that can be incorporated into or communicate with pre-existing active self-
checking
capabilities of existing systems, such that if the system fails, very few
boards could
pass before a problem is discovered.
[0086] An exemplary application of the BoardprintTm technology is described
below in the context of a system used to automatically grade wood pieces 8 in
a
planer mill. The wood pieces 8 are graded boards that vary in thickness and
width
from 2.5 x 10 centimeters to 5 x 30 centimeters (1 x 4 inches to 2 x 12
inches) and
vary in length from 2.4 to 6 meters (8 to 20 feet) long. The boards are kiln
dried and
pass through a grading system during the planning, trimming, and sorting
process.
The grading system determines the board grade based on rules applied to
geometric
and biological characteristics of the boards. Those familiar with the industry
will
know that there are many lumber grade rule sets that depend on the specie and
end
use of the material, and some of these grade rule sets are designated for
products
such as Dimension, Shop, Boards, Timbers, Decking, Stringers, and Machine
Stress-Rated lumber. Several organizations exist that describe and authorize
the
use of different grade rules, and many lumber producers use a combination of
standard grade rules as well as define their own custom grades based on
customer
demand. Two of the most well recognized authorities in this industry are the
American Lumber Standards and the Canadian Lumber Standards.
[0087] The boards travel through a planer at lineal speeds averaging 47
meters/minute (1800 feet/min.), immediately pass through the grading scanner,
and
land on a deck from where chains transport the boards as they are oriented in
a
transverse direction. At this stage in the process, the boards are not
completely
- - CA 02494012 2005-01-24
controlled, and do not necessarily stay in queue. After a few seconds, the
boards
are physically separated into the individual sections of a lugged chain. It is
then
desirable to match a board identity with the information collected at the
grading
scanner because the board will next be trimmed and sorted into a package by
automated equipment. Accordingly, a station 42 is preferably located at a
position
where boards can be identified before they are trimmed and sorted.
[0088] When a board passes through the grading scanner, several small
halogen
lamps can be used to provide lighting for line scan cameras to collect image
data for
the top wide faces 30 and optionally the bottom wide faces 30 of the board.
Skilled
persons will appreciate, however, that preferred embodiments take advantage of
existing multiple-sided scanners at station 12 to collect data from the bottom
and
other surfaces, so that at station 42 bottom face scanning can be omitted.
This
embodiment avoids the costs of adapting and positioning a scanning system to
accommodate the limited openings that the steel structure provides between
chains,
chain races, and supports in order through which a picture of a board bottom
can be
obtained downstream of station 12.
[0089] The board position may be tracked lineally as it travels through the
scan
zone at station 12 using an encoder. However, in preferred embodiments, the
board
is allowed to "float" perpendicularly (sideways) to the direction of travel
because the
image processing is capable of finding the edges of the board and ignoring the
background. This floating allows the system to be more forgiving of miss-
shaped
boards as it does not require the use of side rolls or anvils to guide the
boards. Also,
by not forcing the board to assume a specific shape, preferred embodiments at
stations 12 and 24 allow the boards to be fully relaxed so the normalization
steps
can produce similar results to facilitate a higher likelihood that the board
will indicate
a match when it arrives at station 24.
[0090] The data is collected by one or more of a set of computers 20
containing
hardware and software for controlling the acquisition and data collection
process,
and the board is assigned a sequential board number. The primary use of the
image
data is for grading the board, however, one or more portions of the image data
are
also processed in the computer 20 for its BoardprintTM information. This means
that
the data be warped for size and orientation, FFT values will be generated, the
resulting data will be adjusted to compensate for brightness; and the data
will be
placed in the vector image packet. The image packet is then sent to a host
26
- - CA 02494012 2005-01-24
-
computer 20 where it is stored for the matching process later. This image
packet
contains information for both the top and bottom wide faces 30 of the board,
plus the
sequential board number. The sequence number and image data for this board is
sent to one of the computers 20, each computer 20 preferably containing the
same
hardware and software dedicated to image processing. The next board that is
scanned will be assigned the next number in sequence and its image data is
passed
to another one of the computers 20.
[0091] After the board finally gets singulated onto the lugged
chain, the automatic
processing system controlling the production line activates a camera at the
identification station 42 and an image is acquired of the top face 30 of the
board,
which may be oriented transversely to the direction of travel 6. The position
of board
can be preferably controlled in two ways. The end of the board can be brought
up
against a fixed fence using a set of canted or lobed rolls under the board
before the
board gets to the camera. This even-ending process ensures that the board
image
position is reasonably well guaranteed to be taken at the same position along
the
length of the board for every board. Additionally, an encoder tracks the
lugged chain
and thus the transverse position of the board is known well enough to reliably
be
used to activate the camera to acquire the image. The encoder is physically
connected to the lugged chain and provides signals to the automatic processing
system allowing it to follow any movement of the lugged chain. The lugged
chain
position is therefore known, from which the board position is assumed and used
for
triggering the camera.
[0092] The camera sends the image data via Fire Wire serial format to an
external conversion device where the signal is converted to a fiber optic
medium and
sent to a dedicated board printTM computer 20. When it arrives at the computer
20,
external hardware converts the data back to Fire Wire serial format and stores
the
information in a circular buffer. Software in the computer 20 pulls it out of
the buffer
and processes it as previously described with an additional step of
reorienting the
image. Then, the computer 20 relays the data to the host computer 20 via an
Ethernet connection for the comparison as previously described.
[0093] The host computer 20 compares the vector image packet just received
from the station 42 with the image packets from eleven boards for example,
such as
five boards in the queue ahead of this board, and five boards after this
board.
Skilled persons will appreciate that the finite number in such grouping for
this or
27
- - CA 02494012 2005-01-24
other embodiments may be changed as desirable for statistical purposes,
processing
time constraints, application particulars, or other variables. An exemplary
alternative
number may include ten pieces of wood 8 up and/or downstream, or different
numbers of boards upstream and downstream, such five boards upstream and ten
boards downstream. Comparison steps are employed as previously described
except as modified by the specific embodiment hereinafter described.
[0094] Once the comparison scores are generated, the host computer 20 the host
computer 20 may attempt to match the queue that passed through the scanning
station 42 with the queue that passed through station 12. If the board score
for the
expected match is equal or higher than the scores for the adjacent boards, it
is
considered a match. In some embodiments, the comparison scores between the
board at station 42 and the nearest expected 20 boards in the queue from
station 12
are compared and the highest score determines the most likely order. This
scoring
scheme can result in re-labeling of the order of the queue at station 42. The
technique can cause rejection of board matches that score below a specified
threshold, and can also look backward to reevaluate boards, just in case a
better
sequence becomes evident. This self-correction capability can accommodate most
cases of missing boards as well as boards that break. The matching criteria
can
also be set to reject boards that are introduced at station 42 that have not
passed
through station 12, when the scores are low enough.
[0095] In some exemplary embodiments, if the board score for the expected
match is up to 20% lower than the scores for any adjacent boards, then the
board is
considered a match. However, a probability weighting may require that the next
board have at least a 10% higher score than that of adjacent boards or be
rejected.
If the board score for the expected match is more than 20% lower than scores
for the
adjacent boards, both boards are rejected. Once three boards are rejected in a
row,
the program will automatically adjust to realign the queue. The software does
this
using data from the last three board comparisons to determine whether there is
a
trend for one specific board position that is higher than or equal to the
scores for
adjacent boards as described above. If no trend can be detected in the ten
adjacent
positions, the host computer 20 instructs the automatic processing system to
stop
the material flow so the boards can be manually sorted by an operator.
[0096] In some applications there will be pieces of wood 8 that have not
passed
through station 12 but appear at station 42. In many cases these pieces of
wood 8
28
- - CA 02494012 2005-01-24
-
will be manually graded, so this information is preferably passed to the
BoardprintTM
system so as to reduce the risk that a piece of wood 8 will be rejected, or
possibly
misidentified. The automatic processing system may first learn of such piece
of
wood 8 at some point in the process that may even be after the piece of wood 8
has
already passed through station 42. The automatic processing system preferably
sends such information to the computer 20 and the computer 20 then identifies
the
piece of wood 8 as a special case so its comparison scores will be exempt from
further processing. This capability makes the system much more robust in cases
where a large percentage of pieces of wood 8 regularly bypass station 12.
[0097] Once an identity decision is made, the host computer 20 transmits
the
corresponding sequential number and relevant process information (grade and
trim
decision) to the automatic processing system so it can trim and sort the board
accordingly.
[0098] In some embodiments a piece of wood 8 may be processed, affecting its
length or even its surface before reaching station 42. In some examples, the
processing does not significantly affect grain characteristics, such as
planing or
sanding processes, so that various grain characteristics can still be employed
in the
identification process. In other examples, the piece of wood 8 is trimmed to a
different length between station 12 and station 42. As long as the same areas
of the
piece of wood 8 are used for data capture at both stations and as long as
enough
surface area remains with which to effect a matching identification, the wood
pieces
8 can still be tracked.
[0099] In yet another embodiment, a grading system or the automated
processing
system determines a cutting solution for the pieces of wood 8 that recommends
that
they be cut into multiple smaller pieces of wood. The image characterization
system
collects or sorts image data such that an image packet is created for each of
the
forecast smaller pieces. Skilled persons will appreciate that some embodiments
may
not provide image packets for all forecast pieces of wood 8. An identification
station
can create image packets for the recommended multiple smaller pieces of wood
after they are cut to confirm the identity of some or all of them.
[00100] Skilled persons will appreciate that variations in size, materials,
shape,
form, function, manner of operation, assembly, and use may impact optimum
dimensional and positioning relationships, hardware components, software
applications, and system connectivity.
29
- - CA 02494012 2005-01-24
[00101] It will be obvious to those having skill in the art that many changes
may be
made to the details of the above-described embodiments without departing from
the
underlying principles of the invention. The scope of the present invention
should,
therefore, be determined only by the following claims.
