Note: Descriptions are shown in the official language in which they were submitted.
1323332
Field of the Invention
The present invention relates to color sorting based on
the color of the surface of the piece of wood and where desired
to divide wood pieces having significantly different colors into
smaller elements of the selected different colors and/or
collecting elements of the selected different colors in different
groupings~ The present invention also relates to sorting a piece
of wood based on whether the surface has flat or vertical (edge)
grain appearance and where desired to divide wood pieces having
significantly different grain appearance (flat or vertical) into
smaller elements having substantially only flat or vertical grain
appearance and/or collecting elements of flat and vertical grain
appearance in different groupings.
Background of the Present Invention
In finger jointing of high value lumber elements
particularly those used primarily for decorative purposes, for
example cedar blocks, it is important that the resultant product
be homogeneous both in wood color and slope of the grain to
provide a uniform, pleasing, visual appearance.
The concept of recognizing the color in different areas
of a piece of lu~ber is not new. Such a process has been used
for example in discriminating defects (e.g. knots, dark spots,
etc.) and determining the location of these defects in a piece of
lumber and the lumber then sawn on the basis of the location of
such defects to remove or minimize the effect of the defects in
q~
13233~2
the value of the lumber pieces.
It has also been proposed in Canadian patent 719,067
issued October 5, 1965 to Finlay to use the intensity of
reflected liqht to detect defects (by color) and to cut these
defects from the wood. This patent refers to color sorting but
no means for actual sorting are disclosed and no mechanism for
discerning more than a color representing a defect in the
background of the wood or lumber product is described.
The concept of detecting grain and particularly cross
grain and in some cases torn grain using ultrasonics or
microwaves coupled with light shadow techniques has also been
suggested (see Szymani and MacDonald article entitled 'Defect
Detection in Lumber' State of the Art; Forest Products Journal,
Volume 31, No. 11, November, 1981).
Brief Description of the Present Invention
It is an object of the present invention to provide a
means for discriminating areas of the surface of a wood element
and classify same by color.
It is also an object of the present invention to
provide a means for discriminating areas of the surface of a wood
element and classify same by grain appearance.
Broadly the present invention relates to a method of
sorting wood pieces by color comprising scanning a surface of
said wood piece using a scanning camera, generating separate red
and green (and optionally blue) image data from said camera while
1323332
moving said surface of said wood piece past ~aid camera,
synchronizing said camera with said wood piece to correlate the
acquisition of frames of said red, green and, if used, blue image
data relative to the position on said surface from which it is
generated, each said frame representing one of a plurality of
increments of length along said surface, proces~ing said data
from each of said red and green color images by developing a
histogram of frequency distribution of color intensity for each
said red and green colors for each of said frames and analyzing
said histograms for each of said frames to determine a designated
color for each of said frames, characterizing each said
increments of length based on said designated color for its
respective of said frames and activating a sorter to æegregate
wood pieces of selected colors identified by said analysis into
their selected bins.
Preferably said designated colors are determined based
on the mean values from the histogram for each of said colors.
Preferably means will be provided to separate a given
wood piece into a plurality of pieces based on colors present in
said frames of length by severing said wood piece in the area of
change of color from one selected color to another.
Preferably the camera will be a line scan camera
acquiring data one line at a time transverse to the direction of
movement of the wood piece relative to the camera, and each said
frame will comprise a scan of a plurality of lines covering a
significant increment of length (generally of at least about one
1323332
inch) in the direction of movement of the wood piece relative to
said camera.
In some cases the data will comprise more than one
significant color peak in the histogram for one or more of said
colors in one of said frames and the cut-off means may be
actuated to sever said wood piece in the area of said piece
corresponding to the location of said change within the
respective color image data. The cutting location within a frame
of data representing a discrete length is determined by examining
a profile extending in the direction of travel of the wood piece
relative to the camera for at least one of said colors. Normally
this profile will be along or adjacent the axial centre line of
the wood piece. Where a color change occurs, a profile will
contain two signal levels corresponding to the two intensity
levels representing each color component. The point at which the
one level changes to another will establish the location to be
chopped and will be used to activate the cut-off mechanism.
Prererably each of the histograms will be smoothed to
reduce the noise to signal ratio and then convoluted with a
second derivative Gaussian function to provide a series of zero
crossings which define a color distribution.
Broadly the present invention also relates to a method
of sorting wood pieces by surface grain orientation comprising
scanning a surface of said wood piece using a scanning camera,
generating selected red, green or blue image data in the case of
a color camera or a grey scale image data in the case of a black
1323~32
and white camera from said camera while moving said surface of
said wood piece past said camera, synchronizing said camera with
said wood piece to correlate the acquisition of frames of image
data with the position on said surface from which said frame is
generated, each said frame representing one of a plurality of
increments of length along said surface, processing said selected
image data by developing a histogram of frequency versus
intensity for each said frame and analyzing said histograms for
each of said frame to classify same as representing a surface
having a vertical or flat grain and activating a sorter to
~egregate wood pieces having vertical grain from those having
flat grain.
Preferably the data will be classified as representing
vertical grain or flat grain based on analysis of the freguency
distribution of histogram and determining the frequency
variability of the frequency distribution for each frame and
declarinq those frames as having a frequency variability below a
selected threshold as representing vertical grain and those above
said threshold representing flat grain.
Preferably the camera will be a color camera and said
selected color will be green when a color camera i~ used and
color is being examined.
Preferably means will be provided to separate a given
wood piece into a plurality of pieces based on their
classification of having vertical vs. flat grain by severing said
wood piece in the area of change between flat and vertical grain.
1323332
Preferably the camera for grain detection will be a
line scan camera acquiring data one line at a time transver6e to
the direction of movement of the wood piece relative to the
camera, and each said frame will comprise a scan of a plurality
S of lines covering a significant increment of length (generally
of at least about one inch) in the direction of movement of the
wood piece relative to said camera.
Brief Description of the Drawings
Further features objects and advantages will be evident
from the following detailed description of the preferred
embodiment of the present invention taken in conjunction with the
accompanying drawings in which
Figure 1 is a schematic plan view illustrating a layout
of stations forming the present invention.
Figure 2 is a schematic side elevation illustrating the
layout of the stations of Figure 1.
Figure 3 is a schematic illustration of the data
acquisition and computer control of the color sorter of present
invention.
Figure 4 is a schematic illustration of the processing
of the data for color sorting.
Figure 5 is a typical histogram of a single-colored
discrete length of a surface being analyzed.
Figure 6 is a histogram of a discrete length of a
surface being analyzed and having two distinct colors.
1323332
Figure 7 is a schematic illustration of the data
acquisition and computer control of the present invention.
Figure 8 is a schematic illustration of the processing
of the data for grain sorting.
Figure 9 is a typical histogram of a single-colored
diæcrete length (frame) of a surface being analyzed having flat
grain. ~
Figure 10 is a histogram of a discrete length of a
surface being analyzed having vertical grains.
Description of the Preferred Embodiment
As shown in Figures 1 and 2 the sorter 10 comprises a
main conveyer 12 for carrying the elements 14 in the direction of
the arrow 16 through the various stations of the present
invention. As illustrated schematically the conveyer 12 is
provided with an encoder 18 which registers the position of the
conveyor 12, i.e. by registration of the angular movement of the
roll 20 over which the conveyor belt travels.
The encoder 18 at any given point in time codes the
specific location of any selected point on the conveyor 12 so
that the location of any point on wood pieces 14 travelling with
the conveyor 12 may be determined when the location of a known
point on the wood pieces 14 is established on the conveyor 12.
The sorter 10 incorporates a sensing station 21 that
incorporates a camera 22 that images the surface of the wood
pieces 14. The operation of the camera is synchronized with the
1~23332
movement or location on the conveyor 12 by a suitable sen60r 24
that triggers operations when it senses preferably the leading
edge of each of the pieces 14 or alternatively some specific
marking on each of the pieces 14 to position each of the pieces
relative to the conveyor 12 and thereby provide the necessary
information to accurately position these pieces relative to the
conveyor 12 using the data generated by the encoder 18 and
thereby define any location along the length of the wood piece
14, i.e. distance from point of triggering of the sen~or 24.
The other main elements of the present invention
illustrated in Figures 1 and 2 include a cut-off station 26 and a
sorting station 27. Each of these stations 26 and 27 have their
own conveyor 12C and 12S respectively with encoder 18C and 18S
and sensor 24C and 24S respectively equivalent to the conveyor
12, encoder 18 and sensor 24. The data from all the sensors, i.e.
encoder 18, 18C and 18S, and sensors 24, 24C and 24S is sent to a
tracking section 35 of the main computer control 34 and the
information generated in color sensing station 21 by the sensor
24 and encoder 18 is supplied to the stations 26 and 27 via
computer section 35 and used to control their operations and
define accurately the position of the various pieces 14 on the
conveyors 12C and 12S so that chopping and/or sorting occurs at
the correct location.
The cutoff station 16 is provided with a suitable
cutoff chopper 25 or the like and the sorting station 27
incorporates a plurality of bins 28 each of which in the
1323332
illustrated system has a deflector mechanism 30 actuated to
selectively direct wood pieces into its respective bin 28.
The camera 22 may either be an area scan camera or
preferably will be a line scan camera and preferably will use a
S charge coupled device (CCD) as a sensor for the color image. A
black and white camera is adequate for sensing grain but
generally in a given installation the same camera will be u6ed to
substantially sort by grain and color.
Color Sorting
The data for each of the red, green and blue signals
from camera 22 is digitized in the digitizer 32, there being one
digitizer for each color, the individual digitizer being
designated by 32 followed by the letter R, G and B representing
red, green and blue respectively.
The operation of an area scan camera is slightly
different from that of a line scan camera. However in both cases
the actuation of the data acquisition is continuous during the
movement of the wood pieces 14.
The sensor 24 alerts the main control computer or CPU
34 to activate the triggers 36 for each of the buf~ers 38 there
being one trigger and one buffer 38 for each of the signals and
the corresponding triggers and buffers 38 are each indicated by
the respective numbers of 36 and 38 followed by the initial R, G,
and B to signify red, green and blue respectively. These buffers
38 store the data generated by the camera on command by the CPU
34 which activates the trigger 36 so that the data accumulated in
1323332
each buffer can be correlated with a specific location on the
surface of the wood pieces through the combined position sensing
operation of the encoder 18, sensor 24 and a tracking section 35
in the CPU 34.
Similarly there is an analyzer 40 provided for each of
the red, green and blue signals as indicated by the reference
numerals 40R, 40G and 40s. The data from the buffers 38R, 3BG
and 38B is fed to the analyzers 40R, 40G and 40B reæpectively.
Histograms are generated in the analyzers for each frame of data
and these histograms are processed in the main computer 34 to
operate the cut-off station control 39 which operates the chopper
25 and the sorting station control 44 which operates the various
gates 30 for the bins 23.
The analyzers 40R, 40G, and 40B are essentially the
same and the sequences carried out by each analyzer 40 are
indicated in conjunction with the main computer 34 in Figure 4.
(If desired instead of a plurality of analyzers 40 the image from
the buffers may be multi-plexed one at a time to an analy~er to
produce histograms and these histograms read by the CPU 34 for
color analysis.)
As shown in Figure 4 each frame of input datà from thè
buffer 38 is first histogrammed as indicated at 46. The
histogram produced is a histogram of frequency versus intensity
as detected by each of the elements within the CCD of the camera.
Regardless of the type of camera used the number of discrete
intensity readings in each increment of length (frame) will
1323332
depend on the size of the CCD matrix of the camera (the number
of CCD elements). When using a line scan camera each discrete
length of surface scanned for each histogram generation i.e.
each frame of data processed in the data processing stage when
the line scan camera is used will represent a statistically
significant iength which will generally be at least one inch
measured in the direction of relative movement of the camera and
wood piece (the width will normally correspond to the width of
the wood piece measured perpendicular to the direction of
relative movement).
When an area scan camera is employed, each frame of
data, i.e. each histogram will represent an area of the surface
of the wood piece equivalent in length to the length of the
field of view of the camera in the direction of movement of the
wood pieces (at least one inch) plus the distance the wood piece
travels relative to the camera in the time required to obtain
the frame of data (which normally will be negligible as will be
described below) and the camera 22 and digitizer 38 will be
triggered at discrete intervals to produce frames of data
representing abutting areas of the wood pieces. Obviously if
triggering occurs at regularly spaced intervals and the rate of
relative movement changes the length of travel between
triggering will change. If all of the surface is to be analyzed
or if extra or reprocessing of data is to be avoided triggering
should be correlated with rate of travel.
~`-.
~ 1323332
_ to b~ ~voidcd.
The area viewed by each element of the CCD will
generate a separate intensity and the number of CCD elements
having a given intensity will be represented in the histogram.
The histogram may then be smoothed by convoluting a Gaussian
function with the histogram. A width of 15 and a sigma of 2 for
the Gaussian function have been found satisfactory. Such a
smoothing operation is schematically indicated at 48. The second
pass through the histogram is a convolution of the second
derivative Gaussian as indicated at 50 which will determine a
series of zero crossings such that each adjacent pair of zero
crossings define a distribution (predominant color) and separate
distributions will be determined. The smoothing and convolution
operations 48 and 50 will normally be carried out in the main CPU
34.
The main control computer 34 also determines what the
main color distribution of each histogram is and its main color
component as indicated at 51 in Figure 3.
In most cases there will only be a single color
distribution likely for each of the color channels thereby
indicating a single mean color for the frame of data being
processed. Such curves are illustrated for example in Figure 5
showing one major peak in the distribution for each of the colors
red, green and blue. In this case the area of the wood piece
will be classified based on the mean color for that area and as
long as the mean color of the whole wood element does not deviate
1323332
significantly that whole wood element will be sorted into a
selected bin.
On the other hand, if the color shifts, i.e. the
locations of the red, green or blue peaks or all of them shift
significantly indicating a change in color provided that change
in color is deemed significant based on pre-established color
sorting criteria, the cut-off station 26 is activated at the
appropriate location to cause the chopper 25 to separate the
areas of different colors the location of which in each of the
stations 21, 16 and 27 is defined in the tracking section 35 of
the computer 34. Rnowing the location of the board pieces of
different color permits the sorting control 44 to activate the
deflectors 30 at the appropriate time to direct different colored
pieces into their respective bins 28.
In some cases a single frame of data, i.e. the data
being processed at any one time, will incorporate more than one
predominant color. This will appear in the histogram by a pair
of peaks (see Figure 6) with their distributions being segregated
by the zero crossings determined by the second derivative
Gaussians. This multi-peak or generally two peak distribution
need not be found for all three colors but may be present in only
one color and yet indicate a significant change in color.
The degree of separation between the peaks in a single
histogram (see distance S in Figure 6) and the area they
represent in the image (e.g. hatched area) determines whether the
two colors are high contrast relative to each other and the
14
1323332
percentage of the surface area or frame being 6enfied that is of a
different color, i.e. the significance of the color change. This
information is then used to make a chop decision as to whether or
not a cut should be made in the wood piece correlated using
separate areas of different color.
When there are several different color in a given
frame (Figure 6) its classification may be determined in a number
of ways. The preferred system is to generate a central profile
of one of the colors (preferably green) oriented parallel to the
direction of travel in a profile generator 56 (Figure 4). A
color change in a frame will result in a color profile containing
two signal levels corresponding to the two inten6ity levels
representing each color. The location where one level changes to
another will establish the location for actuation of the chopper
25. The precise location of this change in level can be
determined, for example, by taking the second derivative of such
central color profile as indicated at 58 and basing the chop on
the zero crossing of the second derivative. The calculation of
this chop location based on the sensing of two different colors
can be carried out in the main computer.
The above description has related to analysis of all
three color channels, i.e. red, green and blue, however for many
wood species the color of the wood is represented predominantly
by the green and red colors with blue being almost insignificant
in determining the actual changihg color. With these wood
species it is unnecessary to process the blue channel and thus
1323332
~he decision can be made for many species based on processing of
the red and green signals only.
Grain Sortinq
The description will be related to a color camera
using a selected color as it is the more likely to be used. If
a black and white camera is used grey scale will be used in
place of color intensity.
The data for the selected color which for cedar
preferably will be green (another color may be preferred for
other wood species depending on their surface color and lighting
used) is digitized in the digitizer 32 ~see Figure 7). The
sensor 24 alerts the main control computer or CPU 34 to activate
the trigger 36 for the camera 22 and buffer 38 and correlate the
data for each frame with a specific location on the surface of
the wood pieces through the combined position sensing operation
of the encoder 18, sensor 24 and a tracking section 35 in the
CPU 34. As above described the time of exposure is correlated
with speed so that the image is not significantly blurred.
The data from the buffer 38 is fed to an analyzer 40
that develops a histogram for each frame of data and the
histograms are processed in the main computer 34 to classify the
frame and to operate the cut-off station 26 control 39 whi~h
operates the chopper 25 and the sorting station 27 control 44
which operates the various gates 30 for the bins 28.
-
1323332
The sequence carried out by the analyzer 40 in
conjunction with the main computer 34 is illustrated in Figure 8.
When using a line scan camera each discrete length of
surface scanned (frame) for each histogram generation i.e. each
frame of data processed in the data processing stage will
represent a statistically significant length which will generally
be at least one inch measured in the direction of relative
movement of the camera and wood piece (the width will normally
correspond to the width of the wood piece measured perpendicular
to the direction of relative movement).
When an area scan camera is employed, each frame of
data, i.e. each histogram will represent an area of the surface
of the wood piece equivalent in length to the field of view of
the camera (at least 1 inch) plus the distance the wood piece
travels relative to the camera in the time used to obtain the
frame of data (generally only a very short if blurring is to be
avoided) and the camera 22 and digitizer 38 will be triggered at
discrete intervals to produce frames of data representing
abutting areas of the wood pieces. Obviously if triggering were
to occur at regularly spaced intervals and the rate of relative
movement changes the length travelled between triggering will
change. If all the surface is analyzed or extra or reprocessing
of data is to be avoided triggering should be correlated with
rate of travel.
1323332
As above indicated only one color from the camera 22
need be digitized and analyæed in the analyzer 40 when a color
camera is used. For the remainder of this disclosure it will be
assumed that it is the green color being processed but as above
indicated other colors or grey scale (if a black and white camera
is used) may be selected.
The green image is convoluted using a one dimensional,
first derivative filter directed along the direction of board
travel, i.e. along the length of the board since the board will
be travelling in a direction parallel to its length to derive the
first derivative as indicated at 50 in Figure 4. In some cases
it may be desirable to first smooth the image data using a
suitable averaging kernel to convolute the image and generate an
averaged image before subjecting the data to the one-dimensional
first derivative filter.
The output of the one directional first derivative
filter is histogrammed which represents a histogram of frequency
versus intensity as indicated at 52 and a distribution is
generated about a central mean which for most cases will be 128
(the twos complement of an 8 bit image). The resultant
distribution is biased and contains negative components due to
the first derivative calculation so that above the mean is in a
positive range and below the mean is a negative range, i.e. for
each histogram based on the central mean of 128 the mean of each
histogram will always be at the selected 128 value and the degree
of variability and distribution from this mean of 128 determines
18
13233~2
whether the board will be classified as having vertical or flat
grain.
The frequency about the mean as indicated at 54 in
Figure 8 and the degrees of frequency variation is determined as
indicated at 56 in Figure 8.
It will be apparent that the first derivative obtained
at 50 calculates the edge components sensed axially along the
length of the board. In the case of a flat grain there are many
edges that are generated and the first derivative will enhance
those edges and subsequently generate a distribution with a high
variability. A vertical grain board on the other hand will
generally have the grain extending lengthwise of the board, i.e.
in the direction of travel of the board which in turn is matched
with the direction of the filter and hence will not generate any
significant amount of edge response resulting in a frequency
distribution about the mean that will have very low variability.
To discriminate flat grain from vertical grain a
threshold value T :is selected and boards exhibiting a degree of
frequency variabilit~ greater than the selected T will
automatically be classified as flat grain (see VF in Figure 91
and those with a variability of less than the selected threshold
value T (see Vv in Figure 10) will be designated as vertical
grain. Th0 value T will be determined empirically for any
species.
The cutoff station control 39 will be activated when
the classification of the grain from one frame is different from
19
1323332
that of the following frame to cut the board at the junction
between the two frames which as above indicated is e~tablished by
the encoders 18 and 18C operating in conjunction with the sensors
24 and 24C and the tracking control 35.
Similarly the sorting station control 44 will be
actuated to direct pieces of board having vertical grain to one
sort bin 28 and those having flat grain to a second bin location
of these pieces of different grain being coordinated via the
sensors 24 and 24S and the encoders 18 and 18S so the location of
each board section is known throughout the equipment and the
sorting bins be actuated accordingly.
Combined Sorting
It will be apparent that the same scanning and
positioning equipment may be used for color sorting and g{ain
sorting by handling the information in two different ways as
described above for grain and color sorting and combining the
results to chop and sort the wood in accordance with both grain
and color.
Having described the invention modifications will be
evident to those skilled in the art without departing from the
spirit of the invention as defined in the appended claims.
