Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-1-
MEAT IMAGING SYSTEM FOR PALATABILITYAND YIELD PREDICTION
1. Field of the Invention
The field of the present invention is prediction of meat palatability and
yield. More
specifically, the present invention relates to the prediction of meat
palatability and yield by use
s of image analysis ( IA) to determine the color parameters L* (psychometric
lightness), a* (red
vs. green), and b* (yellow vs. blue) or the tissue density of the lean and fat
portions of a meat
animal carcass or cut.
2. Description of Related Art
Consumers of meat generally prefer, and are willing to pay for, greater meat
tenderness.
io Marbling score of a carcass has been shown to generally correlate with
subsequent cooked meat
palatability across a wide range of marbling levels for beef, pork, and lamb.
However, between
carcasses with the same marbling level, there are substantial differences in
palatability. Other
factors of the carcass believed to predict palatability include maturity
score, muscle pH, and
muscle color; these factors may be more valuable in the prediction of
palatability of chicken,
is turkey, and fish. Among those with expertise in carcass examination, e.g.
meat scientists and
U.S. Department of Agriculture (USDA) graders, some of these factors can be
scored and
palatability predicted by assigning a USDA Quality Grade, given sufficient
examination time.
In practice, for the example of beef, USDA graders working at packing plants
commonly must
assign Grades to 250 to 450 beef carcasses per hour, which does not provide
enough time for a
ao complete examination of all factors related to prediction of palatability.
The shortage of time
also makes difficult the required accurate computation of Quality Grades.
In addition, USDA graders are required to compute Yield Grades, which are
intended to
estimate the cutability and composition of a carcass. Factors used to
determine Yield Grades
include hot carcass weight, ribeye area (cross-sectional area of the
longissimus m. at the 12-13th
Zs rib interface), estimated kidney, pelvic, and heart fat percentage, and
actual and adjusted
subcutaneous fat thickness at the carcass exterior. The time constraints
described above for the
calculation of Quality Grades also apply to the calculation of Yield Grades.
The parameters that
underlie the assignment of Quality Grades and Yield Grades are published by
the USDA
Agricultural Marketing Service, Livestock and Seed Division, e.g., for beef,
the United States
so Standards for Grades of Carcass Beef.
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-2-
A device for scoring factors predictive of palatability of a meat carcass or
cut, in addition
to an examination of the carcass or cut by a USDA grader would allow meat
palatability to be
more accurately predicted and USDA Quality Grades to be more accurately
assigned. This
would allow greater consumer confidence in the Quality Grading system, as well
as any
s additional system for certification of conformance to product quality
specifications, as would be
desired in a "brand-name" program. In either event, more precise sortation of
carcasses for
determining meat prices would be allowed. This superior sortation would
provide economic
benefit to those at all segments of the meat production system: restaurateurs,
foodservice
operators, and retailers; packers; feed lot operators; and ranchers, farmers,
and harvesters of
~o pork, lamb, beef and dairy cattle, chicken, turkey, and various fish
species. This superior
sortation would also benefit scientists in the collection of carcass and cut
data for research, and
the previous owners of livestock in making genetic and other management
decisions.
Several attempts have been made to construct such devices for use in the beef
industry.
One such device uses a "duo-scan" or "dual-component" image analysis system.
Two cameras
is are used; a first camera on the slaughter floor scans an entire carcass,
and a second camera scans
the ribeye after the carcass is chilled and ribbed for quartering. In the use
of these systems,
video data are recorded from a beef carcass and transferred to a computer. A
program run by the
computer determines the percentages of the carcass comprised of fat and lean
from the recorded
image and additional data available, e.g. hot carcass weight. The quantities
of cuts at various
Zo levels of lean that can be derived from the carcass are then predicted.
However, based on
scientific evaluation, the system is not able to predict palatability of the
observed carcass for
augmenting the assignment of a USDA Quality Grade or other purpose related to
sorting
carcasses based on eating quality.
One possible set of factors that can be examined to predict palatability is
muscle and fat
Zs color. Wulf et al., J. Anim. Sci. (1997) 75, 684, reported results of both
color scoring in the
L*a*b* color space of raw longissimus thoracis muscle at 27 h postmortem, and
Warner-
Bratzler shear force determinations of aged, thawed, cooked longissimus
lumborum muscle,
from carcasses of cattle derived from crosses between various breeds of Bos
taurus (European-
based genetics) and Bos indicus (heat-tolerant, tropically-based genetics).
Tenderness, as
3o measured by shear force, correlated with all three color measurements, with
the highest
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-3-
correlation seen with b* values. These results demonstrated that muscle color
can be used to
predict beef palatability.
Among other factors that can be examined to predict palatability are lean
tissue density,
fat tissue density and connective tissue density. Park et al., J. Food. Sci.
(1994) 59:697-701,
s reported results of A-mode (one-dimensional brightness) ultrasonic spectral
feature analysis.
Tenderness correlated with resonant frequency, juiciness and flavor correlated
with the number
of local maxima. These results demonstrated that ultrasonic spectral features
and other methods
known in the art for determining tissue density can be used to predict beef
palatability.
Therefore, it is desirable to have an apparatus for scoring factors predictive
of the
io palatability of a meat animal carcass. It is desirable for such an
apparatus to collect and process
data and provide output within the time frame that a carcass is examined by a
USDA grader
under typical conditions in the packing house, commonly 5-15 sec. It is
desirable for such an
apparatus to return scores for at least one of, for example, color and color
variability of lean
tissue, color and color variability of fat tissue, extent of marbling, average
number and variance
is of marbling flecks per unit area, average size of marbling and the variance
of average marbling
size, average texture, firmness of lean tissue, lean tissue density, fat
tissue density and
connective tissue density. It is desirable for the apparatus to use these
measures to assign a
grade or a score to carcasses in order that the carcasses can be sorted into
groups that reflect
accurate differences in cooked meat palatability. It is also desirable for the
apparatus to use
Zo these measures to identify defect conditions in the meat such as, but not
limited to, bruising, dark
cutter or heat ring.
It is also desirable to have an apparatus for measuring the cross-sectional
surface area of
an exposed, cut muscle (e.g. ribeye) for use in predicting the composition
(fat, lean, bone) of a
carcass or cut. It is desirable for the apparatus to use this measure to
assign a grade or score to
Zs carcasses in order that the carcasses can be sorted into groups that
reflect accurate differences in
yield. It is desirable for this apparatus to also measure relative areas of
cross-section surfaces
comprised of fat and/or bone. In addition, it is desirable to have an
apparatus for measuring,
predicting, and sorting carcasses on the bases of any combinations of
palatability, defect
conditions, and yield.
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-4-
Further, it is desirable for such an apparatus to be portable, e.g. small and
lightweight. It
is desirable for the apparatus to be capable of withstanding packing plant
environments, e.g. to
be mounted in a protective housing.
3. Summary of Invention
s The present invention is related to a method for predicting the palatability
of meat,
comprising: providing image data related to at least a portion of the meat;
analyzing the image
data to distinguish at least one area of interest of the meat; analyzing the
image data
corresponding to each area of interest to measure at least one characteristic
of the area of
interest based on the image data; predicting the palatability of the meat
based on the
io characteristic.
The present invention is also related to an apparatus for predicting the
palatability of
meat, comprising: an imaging device adapted to provide an image data of at
least a portion of
the meat; a data processing unit adapted to execute program instructions; a
program storage
device encoded with program instructions that, when executed, perform a method
for predicting
is the palatability of meat, the method comprising: analyzing the image data
to distinguish at least
one area of interest of the meat; analyzing the image data corresponding to
the area of interest
to measure at least one characteristic of the lean section based on the image
data; and predicting
the palatability of the meat based on the characteristic.
FIG. 1 shows a schematic view of an apparatus of the present invention.
ao FIG. 2 shows a flowchart of a method of the present invention.
FIG. 3 shows a flowchart of a computer program analyzing image data to
distinguish at
least one area of interest of the meat, analyzing the image data corresponding
to the area of
interest to measure at least one characteristic of the area of interest based
on the image data.
The present invention provides an image analysis ( IA) system for scoring
factors
Zs predictive of the palatability of a meat animal carcass. The IA system may
be any type of
imaging system known to those of skill in the art, such as a camera,
tomographic imaging,
magnetic resonance imaging (MRI), sound wave imaging, radio wave imaging,
microwave
imaging, or particle beam imaging, and is preferably a color video IA system.
As shown in
FIG. 1, the IA system includes an imaging device 12, preferably a 3- CCD color
video camera,
so preferably mounted in an enclosurell. The imaging device 12 optionally
features an
illumination system 26 mounted either on the imaging device, on the imaging
device enclosure,
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-5-
or not on the imaging device but in the imaging device enclosure. The
illumination system 26
may be any light emitting device known to those of skill in the art, or a
source of energy of
equivalent function meant to impinge on the sample of meat for measurement in
the parts of the
energy spectrum corresponding to the sensitivity ranges required by the type
of imaging device
s 12, including visible light, infrared light, ultraviolet light, x-rays,
gamma rays, electrons,
positrons, electrical fields, magnetic fields, sonic wave, ultrasonic wave,
infrasonic wave or
microwaves. The IA system also includes a data processing unit 16, the data
processing unit 16
interfaced with a program storage device 20 by a program storage device
interface 18, and at
least one output device 24 by an output device interface 22.
to The program storage device 20 contains a computer program or programs
required for
proper processing of image data, preferably color video image data, by the
data processing unit
16. The data processing unit 16 is linked to, and receives data from, the
video camera 12 via
either a transfer cable 14 or a wireless transmission device (not shown). The
data processing
unit 16 comprises a standard central processing unit (CPU), and, where
necessary or appropriate,
~s preferably also a software module or hardware device for conversion of
analog data to digital
data, and processes image data according to instructions encoded by a computer
program stored
in the program storage device 20. Image data can be used in subsequent
calculation of the values
of characteristics, the values being predictive of palatability, the
characteristics including color
and color variability of lean tissue, color and color variability of fat
tissue, extent of marbling,
zo average number and variance of marbling flecks per unit area, average size
of marbling and the
variance of average marbling size, average texture of marbling and lean
tissue, firmness of lean
tissue, density of lean tissue, density of fat tissue, and density of
connective tissue. These values
can then be used to sort meat (herein defined as a meat animal carcass, side,
or cut, or any
portion of a carcass, side, or cut) into groups that vary in predicted
subsequent cooked eating
as quality.
The color parameters L*, a*, and b*, or the tissue density parameters can also
be used to
calculate the values of factors predictive of yield, such as the cross-
sectional area of a muscle of
interest and other surrounding organs such as fat, bone, and connective
tissue. These values can
then be used to sort meat into groups that vary in predicted composition.
so The color parameters L*, a*, and b* can also be used to calculate the
values of factors
predictive of defect conditions of the meat, such as bruising, dark cutter and
heat ring. These
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-6-
values can then be used to denote the carcass as defective or to adjust or
otherwise alter the
quality sortation decision.
The data processing unit 16 is linked to, and transmits results of data
processing to, at
least one output device 24 by output device interface 22. Optionally, results
of data processing
s can also be written to a file in the program storage device 20 via program
storage device
interface 18. An output device 24 can be a video screen, printer, or other
device. It is preferred
that at least one output device 24 provide a physical or electronic tag to
label the meat 10 with
results of data processing, in order to facilitate sortation of meat animal
carcasses, cuts, or both
into groups with similar predicted palatability and/or yield.
io The present invention also provides a method of predicting the palatability
of the meat 10
and determining the cross-sectional area of the meat 10. Using the color IA
system referred to
above, image data collected from meat 10 is recorded by the imaging device 12,
processed by
the data processing unit 16, and the values of palatability and/or muscle
cross-sectional area is
output by the output device 24 to augment the observations made by a USDA line
grader, or
i s other operator responsible for sorting or characterizing meat animal
carcasses, in order to allow
more accurate assignment of Quality Grades, Yield Grades, defect conditions,
and/or other
sorting or classification criteria based on the characteristics.
An apparatus for use in the present invention comprises an imaging device 12
and a data
processing unit 16. The imaging device 12 can be any such imaging device known
to those of
Zo skill in the art, such as a camera, tomographic imaging (i.e. CAT, PET)
device, magnetic
resonance imaging (MRI) device, sound wave imaging device, radio wave imaging
device,
microwave imaging device, or particle beam imaging device. If the imaging
device is a camera,
the camera is at least one of a photographic camera, a digital still camera,
and a video camera.
The camera responds to light in at least one portion of the light spectrum,
each such portion
as consisting of a band of, such as a segment of ultraviolet wavelengths (200
nm to 400 nm),
visible wavelengths (400nm to 700nm), infrared wavelengths (700 nm to 10 m),
or po r tions
thereof.
If the image is tomographic, it can be obtained by at least one of x-ray
tomography, and
particle beam tomography, such as computer axial tomography (CAT) or positron
emission
so tomography (PET). These devices produce non-invasive cross-sectional and
three-dimensional
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
images resulting from the transmission and scattering characteristics of the
specimen to the type
of incident energy, where intensity is a function of object cross-sectional
density.
If the image is a sound wave image, it can be obtained by ultrasound, B-mode
ultrasound, or infrasonic imaging. These devices produce non-invasive cross-
sectional and three-
s dimensional images resulting from the transmission and reflection
characteristics of the
specimen to the frequency of sound wave applied, where intensity is a function
of object cross-
sectional density. -
It is important for the imaging device 12 to provide output within the time
frame allotted
for meat carcass examination, typically 5-15 seconds. Preferably the output is
in real-time.
io Such real-time output can be the same technology as the viewfinder on a
known camcorder or
video camera, the real-time output can be the same technology as a known
digital camcorder, the
real- time output can be a known computer-generated real-time display as are
known in video-
conferencing applications, or can be any other technology known to those of
skill in the art. It is
preferable for the imaging device 12 to be a color video camera, for reasons
discussed below. It
is is also preferred that the imaging device 12 be small and lightweight, to
provide the advantages
of portability and flexibility of positioning, i.e. adjusting the camera angle
by the user to provide
for optimal collection of image data from the meat 10. It is also preferred
the imaging device 12
be durable, in order to better withstand the environment of the packing plant.
The power source
of the imaging device 12 can be either direct current, i.e. a battery secured
to electrical contacts
Zo from which the imaging device 12 can draw power, or alternating current
provided from either
an electrical outlet or from the data processing unit 16.
An illumination system 26 can optionally be used to illuminate the meat with
energy in
the useful spectrum of the imaging device. This is desirable when using
visible light imaging
and the ambient lighting is poor or uneven or when it is desired to examine
regions of the meat
as 10 that are not illuminated by ambient light, or when the spectral
sensitivity of the imaging
device is not in the visible light part of the electromagnetic spectrum. Any
known or future
developed illumination system 26 can be used, e.g. a lamp (incandescent,
fluorescent, etc.), a
laser, etc. for visible and near-visible portions of the spectrum , x-rays,
gamma rays, electrons,
electrical fields, magnetic fields, sonic beam, ultrasonic beam, infrasonic
beam,or microwaves.
3o The power source 42 of the illumination system 26 can be either direct
current, i.e. a battery 45,
or alternating current drawn from either an electrical outlet 50, the imaging
device 12, or the
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
_g_
data processing unit 16. It is preferred that the illumination system 26 be
small and lightweight,
for reasons discussed in reference to the imaging device 12, above. The
illumination system 26
can be mounted on the imaging device 12, on the outer surface of an imaging
device enclosure
11, or within an imaging device enclosure 11, which is described hereafter.
The imaging device 12 and optional illumination system 26 can be unenclosed or
enclosed. Preferably, the imaging device 12 is enclosed in an imaging device
enclosure 11 for
protection against the environment of packing and processing plants. It is
important for the
imaging device enclosure 11 to provide a first aperture 13 for the lens of the
imaging device 12
to observe the meat 10. If an optional illumination system 26 is used, the
illumination system 26
io can be mounted either on the outer surface of the imaging device enclosure
11 or within the
imaging device enclosure 11. If mounted within the imaging device enclosure
11, the
illumination system 26 can be mounted on the imaging device 12. If the
illumination system 26
is mounted in the imaging device enclosure 11, it is important for an aperture
to be provided for
illumination of the meat 10, either the first aperture 13 used by the lens of
the imaging device 12
is or a second aperture. In either case, the aperture can be unencased or it
can be encased by a pane
of a transparent material 14, wherein "transparent" is defined as allowing the
passage of
substantially all of the energy type and wavelength emitted by the
illumination system 26 or
detectable by the imaging device 12.
If image data is to be transferred from the imaging device 12 to the data
processing unit
Zo 16 by a transfer cable 44 connected therebetween, it is also important for
the imaging device
enclosure 11 to provide an aperture 31 for the cable to exit the enclosure.
This aperture can be
the first aperture 13 used by the lens of the imaging device 12, the second
aperture that can be
used by the illumination system 26, or a third aperture 31. If the cable exits
the enclosure from
the first or second aperture, and the first or second aperture is encased by a
pane of transparent
is material 14, it is important to provide a first cable-passage aperture in
the pane for passage of the
cable. It is preferred that the imaging device enclosure 11 be constructed
from a lightweight
material and be only large enough to conveniently fit the imaging device 12,
and optionally the
illumination system 26 described above.
If alternating current is to be used as the power source of the imaging device
12, it is
3o important for an aperture to be provided to pass the power cable from the
imaging device 12 to
the power source. Any one of the first, second, or third apertures can be
used, or a fourth
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-9-
aperture can be used. If the aperture to be used is encased by a pane of
transparent material, it is
important to provide a second cable-passage aperture in the pane for passage
of the power cable.
Alternatively, both the power cable and the data-transfer cable can exit the
imaging device
enclosure through a single cable-passage aperture.
Optionally, the imaging device enclosure can be designed with features to more
readily
allow user grip and manipulation, e.g. handles, helmet mounting, etc., and/or
with features to
allow fixing in position without user grip and manipulation, e.g. brackets for
wall mounting,
ceiling mounting, or tripod mounting, among other features. Optionally, wall,
ceiling, or tripod
mounting can be to motorized rotatable heads for adjusting imaging device
angle and focal
io length.
Preferably, the imaging device enclosure can be designed to be easily opened
to allow for
convenient maintenance of the imaging device 12 or replacement of a battery if
direct current is
used as the power source of the imaging device 12. Maintenance of the
illumination system 26
may also be needed, and preferably in this option will be allowed by the same
easy-opening
is design described for the imaging device 12. The easy-opening design can be
affected by the use
of screws, clamps, or other means widely known in the art. Ease of maintenance
is desirable to
minimize any downtime that may be encountered.
After image data is captured by the imaging device 12, it is transferred in
real- time to
the data processing unit 16. Data can be transferred by a transfer cable 14 or
by a wireless data
Zo transmission device (not shown). In most situations, transfer cable 14 is
the preferred medium
of transmission based on superior shielding and lower cost. In situations
where the imaging
device 12 and data processing unit 16 are widely separated, a wireless data
transmission device
(not shown) can be a more practical medium of transmission. Any technique of
data transfer
presently known or developed in the future by those of skill in the art can be
used.
Zs The video image data can be sent from the video camera 12 to the data
processing unit 16
as either analog or digital data. If sent as analog data, it is important to
convert the data to
digital data before processing by sending the data to a hardware device (not
shown) or software
module capable of converting the data. Such a hardware module may be termed a
"video frame-
grabber". If the image data is sent as digital data, no conversion is required
before processing
3o the data.
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-10-
For purposes of the present invention, a "data processing unit" 16 is defined
as including,
but not limited to, desktop computers, laptop computers, handheld computers,
and dedicated
electronic devices. Any data processing unit presently known or developed in
the future by
those of skill in the art can be used in the present invention. In one
embodiment of the present
s invention, the data processing unit 16 can be small and lightweight to
provide portability. In a
second embodiment of the present invention, the data processing unit 16 can be
a
microcomputer, minicomputer, or mainframe that is not portable. The present
invention is not
limited to any specific data processing unit, computer, or operating system.
An exemplary
embodiment, but one not to be construed as limiting, is a PC- compatible
computer running an
io operating system such as DOS, Windows, or UNIX. The choice of hardware
device or software
module for conversion of analog data to digital data for use in the present
invention is dependent
on the imaging device 12, data processing unit 16, and operating system used,
but given these
constraints the choice will be readily made by one of skill in the art.
Where the imaging device 12 is a visible light sensitive device such as a
color camera, it
is is also preferred that the data processing unit 16 comprises a software
module that converts RGB
color to L*a*b* color. An exemplary software module for this purpose is sold
in HunterLab
Color Vision Systems (Hunter Associates Laboratory, Inc.).
In addition to a cable port 46 or a wireless data transmission device (not
shown) to
receive data from the imaging device 12, it is also preferred that the data
processing unit 16
Zo include other input devices, e.g. a keyboard, a mouse or trackball, a
lightpen, a touchscreen, a
stylus, a bar code reader, etc., to allow convenient exercise of user options
in camera and
software operation, data processing, data storage, program output, etc.
There are several pieces of software (not shown) which it is important for the
data
processing unit 16 to store in a program storage device 20 (examples of
program storage devices
Zs being a hard drive, a floppy disk drive, a tape drive, a ROM, and a CD-ROM,
among others),
access from the program storage device 20 via program storage device interface
18, and execute.
It is important for the data processing unit 16 to have an operating system,
and any necessary
software drivers to properly control and retrieve data from the imaging device
12 and send
output to the at least one output device 24. It is important for the data
processing unit 16 to
3o execute a program or programs that can process received image data,
calculate various
parameters of the muscle imaged in the received image data, and output the
results of the
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-11-
calculations to an output device 24. An exemplary code for such a program or
programs is given
in an appendix hereto. An exemplary flowchart for such a program or programs
is given as FIG.
3.
Where the imaging device 12 is a visible light device such as a color camera,
the image
s data can be analyzed for color scale parameters. If it is desired to conform
to international
standard, the image data can be analyzed for the color scale parameters L*,
a*, and b*, as
defined by the Commission Internationale d'Eclairage (CIE). A set of L*a*b*
parameters is
recorded for each frame. L*, a*, and b* are dimensions of a three-dimensional
color space
which is standardized to reflect how color is perceived by humans. The L*
dimension
~o corresponds to lightness (a value of zero being black, a value of 100 being
white), the a*
dimension corresponds to relative levels of green and red (a negative value
being green, a
positive value being red), and the b* dimension corresponds to relative levels
of blue and yellow
(a negative value being blue, a positive value being yellow). In a preferred
embodiment, the
system can capture pixelated images from areas of 12 to 432 square inches (75
to 2700 cm2)
~s from the muscle of interest, comprising up to 350,000 pixels per
measurement, and determine
L*, a*, and b* for sub-regions of the image frame comprising at least one
pixel each. In all
embodiments, it is desirable for determination of L*a*b* to be performed using
the HunterLab
software conversion or similar module. Once the value of L*a*b* is determined,
at least one of
the L*, a*, and b* components can be used in subsequent data processing.
Zo The image data can be further analyzed to identify areas of interest
corresponding to
factors predictive of palatability using any such methods of analysis known to
those of skill in
the art, including, among others, image segmentation, histogram thresholding,
spatial analysis,
pattern matching, pattern analysis, neural network, region growing, and focus
of attention
methods, as described in numerous references, such as The Image Processing
Handbook 3'a
Zs Edition, 1999, John C. Russ, CRC Press.
In the preferred embodiment, after determination of L*, a*, and b* for each
area of
interest, a program then calculates several parameters of the image for each
frame. First, the
program outlines the muscle of interest by choosing areas that have tolerances
of b* compatible
with muscle. A sorting of at least one area of the image into one of two
classifications, as in
3o muscle and non-muscle, may be termed a "binary mask." Areas with values of
b* compatible
with the muscle of interest are then examined for their L* and a* scores for
verification and
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
- 12-
rejection of surrounding tissues invading the outline of the muscle of
interest. Further
examination need not be performed on areas with L*, a*, and b* scores
suggestive of bone,
connective tissue, and fat. The surface area of the cross-section of the
muscle of interest is
determined.
Within the portion of the image taken from the muscle of interest, the lean
tissue and fat
tissue of the muscle can be distinguished and raw L*, a*, and b* scores, and
variation in L*, a*
and b* scores across the area of interest, for the lean tissues of the muscle
can be determined.
These scores can then be sent to the output device 24 to be displayed in
numerical format and/or
retained to calculate quality- and yield-determining characteristics as
described below. It is
io known that, among other characteristics, higher values of b* for lean
tissues of muscle correlate
with greater tenderness (Wulf et. al., 1996). In addition, the fat color and
color variability of
intermuscular fat can also be determined.
Also within the portion of the image taken from the muscle of interest,
determinations
can be made of the quantity, distribution, dispersion, texture, and firmness
of marbling
is (intramuscular fat deposited within the muscle). The quantity of marbling
can be determined by
calculating the percentage of muscle surface area with L*, a*, and b* scores
compatible with fat
tissue.
In addition to calculating the quantity of marbling present, the distribution
and dispersion
of marbling can be determined. First, the portion of the image derived from
the muscle of
z.o interest can be divided into subcells of equal size. A size of 64 x 48
pixels can be used. Within
each subcell, the number of marbling flecks can be determined as the number of
discrete regions
with L*, a*, and b* values corresponding to fat, and the average number of
marbling flecks per
subcell can be calculated. The variance of numbers of marbling flecks across
all subcells can be
calculated as well.
zs In addition, the average size of each marbling fleck can be determined
throughout the
muscle of interest from the number of pixels within each discrete region with
L*, a*, and b*
values corresponding to fat. The variance of marbling size across all marbling
flecks can be
calculated as well. The texture and fineness of marbling can also be measured.
It is well known
that generally, greater amounts of more uniformly distributed and finer-
textured marbling reflect
so a higher marbling score and thus meat of higher eating quality.
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-13-
Also, the program can use L*, a*, and b* data to calculate the average
texture, i.e. cross-
sectional surface roughness, of the muscle, and also the firmness of the lean
tissue of the cross-
sectional muscle. It is well known that the surface roughness of a muscle is
inversely correlated
with tenderness, and greater firmness is correlated with flavorfulness.
s In systems wherein the imaging device 12 is a sound wave imager, tomographic
imager,
magnetic resonance imager, radio wave imager, microwave imager, or particle
beam imager, the
program can determine the density of the lean tissue, the density of the fat
tissue and the density
of the connective tissue.
To summarize, characteristics of the area of interest of the meat 10 that can
be measured
to include, but are not limited to, the color of the lean tissue, the color
variation of the lean tissue,
the color of fat tissue, the color variation of the fat tissue, a marbling
quantity, a marbling
distribution, a marbling dispersion, a marbling texture, a marbling fineness,
an average texture of
the lean tissue, a firmness of the lean tissue, a surface area of the lean
section, the density of the
lean tissue, the density of the fat tissue and the density of the connective
tissue. Quantities of the
is non-lean section of the meat 10, including but not limited to the color of
fat, the density of the
lean tissue, the density of the fat tissue, the density of the connective
tissue and the relative areas
of cross-section surfaces comprised of fat, bone, and/or connective tissue,
may be calculated as
well. Other characteristics that one of skill in the art of meat science can
readily see may be
calculated from the values of L*, a*, and b* and that can be predictive of
palatability can be
Zo calculated by the program, and any such characteristics are considered to
be within the scope of
the present invention.
Once values of the several parameters described above are calculated, the
program can
output to the output device 24 calculated values of any or all of the
characteristics given above:
color of lean tissue, color variation of lean tissue, color of fat tissue,
color variation of fat tissue,
Zs extent of marbling, average number of marbling flecks per unit area,
variance of marbling flecks
per unit area, average size of marbling, variance of the average size of
marbling, texture and
fineness of marbling, average texture of lean tissue, firmness of lean tissue,
the density of the
lean tissue, the density of the fat tissue and the density of the connective
tissue. Preferably, the
calculated values of the characteristics, if output, are displayed as
alphanumeric characters that
3o can be conveniently read by the operator. Alternatively, or in addition, to
outputting values of
characteristics to an output device 24, further calculations can be performed
using at least one of
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-14-
the values, and optionally values of parameters input by the operator, to
derive estimated Quality
Grades or other overall indices of cooked meat palatability, which can then be
output.
Further, other parameters not determinable by means of analyzing the image
data may be
used to augment the characteristics determined from the image data, including,
but not limited
s to, pH, connective tissue quantity, connective tissue solubility, sarcomere
length, protease
enzymatic activity, calcium measure, electrical impedance, electrical
conductivity, and tissue
density. These additional characteristics of the meat may be used singly or in
combination to
improve the accuracy of prediction of palatability over that obtained solely
using characteristics
measured from the image data.
io In addition, because a specific muscle of interest has been isolated in a
cross-sectional
image, and the geometry and distance of the apparatus relative to the meat 10
can be known, the
area of the cross-sectional surface of the muscle portion of the meat 10 can
be calculated and
output to the output device 24. Alternatively, or in addition, to outputting
the area to an output
device 24, further calculations can be performed using the area of the cross-
sectional surface of
is the muscle, other parameters readily seen by one of skill in the art of
meat science as calculable
from the L*a*b* data, other image-derived parameters, or values of parameters
input by the
operator, to derive estimated Yield Grades or other overall indices of
composition of the meat
10.
Further, other parameters not determinable by means of analyzing the image
data may be
Zo used to augment the characteristics determined from the image data,
including, but not limited
to, subcutaneous fat depth. These additional characteristics of the meat may
be used singly or in
combination to improve the accuracy of prediction of Yield Grades over that
obtained solely
using characteristics measured from the image data.
The results reported by the program can be output to any output device 24,
such as a
Zs screen, printer, speaker, etc. If operator evaluation of the results is
desired, results can
preferably be displayed on a screen. Preferably, the screen is readily visible
to the grader,
evaluator, or operator at his or her stand. Alternatively, or in addition, it
is preferable that results
be printed or output in such a manner that the outputted- results can be
transferred and affixed to
the meat 10. The manner for outputting results can be text, symbols, or icons
readable by
so personnel either in the packing plant or at later points in the meat
production system.
Alternatively, the manner for outputting results can be a barcode or other
object that can be read
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-15-
by appropriate equipment and decoded into forms readable by personnel at
various points in the
production system. Output results can be affixed to the meat 10 by methods
well-known to the
art, which include, but are not limited to, pins, tacks, and adhesive.
The power source 50 of the data processing unit 16 can be either direct
current, i.e. a
s battery, or alternating current drawn from an electrical outlet.
In the embodiment wherein the data processing unit 16 is dedicated for use in
the present
apparatus, the data processing unit 16 can be mounted in a data processing
unit enclosure 15 or
in the imaging device enclosure 11, or can be unenclosed. In the embodiment
wherein the data
processing unit 16 is a microcomputer, minicomputer, or mainframe computing
resource present
~o in the plant or facility where the apparatus is used, enclosure is not
required. In the embodiment
wherein the data processing unit 16 is a separate, stand-alone, portable
entity, preferably the data
processing unit 16 is mounted in a data processing unit enclosure 15.
It is important for the data processing unit enclosure to provide an aperture
47 or
apertures for output of data to or display of data by the output device 24.
For example, if display
is is to be performed using a video screen congruent with the data processing
unit 16, it is
important for the data processing unit enclosure to provide an aperture 47 for
observation of the
video screen therethrough. Such an aperture can be unencased or it can be
encased by a pane of
transparent material 48, such as glass, plastic, etc. If display is to be
performed by an external
device, e.g. a remote monitor or printer, it is important for the data
processing unit enclosure to
Zo provide an aperture (not shown) for passage of an output cable 22
therethrough. If the data
processing unit 16 is powered by alternating current, it is important for the
data processing unit
enclosure 15 to provide an aperture 49 for passage of a power cable
therethrough. If it is desired
to store outputs to an internal floppy disk drive (not shown), it is important
for the data
processing unit enclosure 15 to provide an aperture (not shown) for insertion
and removal of
Zs floppy disks into and from the internal floppy disk drive therethrough. If
it is desired to store
outputs to an external program storage device 20, it is important for data
processing unit
enclosure 15 to provide an aperture 19 for passage of a data-transfer cable 18
therethrough.
Preferably, if the data processing unit 16 is a dedicated stand-alone unit,
the data
processing unit enclosure 15 is only large enough to conveniently fit the data
processing unit 16,
3o and is lightweight. Optionally, the data processing unit enclosure 15 can
be designed with
features to more readily allow user manipulation, e.g. handles. In this
embodiment, it is also
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-16-
preferred that the data processing unit enclosure 15 be amenable to easy
opening to allow for
convenient maintenance of the data processing unit 16. The easy-opening design
can be affected
by means described for the camera enclosure supra.
The apparatus described above can be used in methods for predicting the
palatability,
s yield, and/or defect conditions of, or augmenting the assignment of USDA or
other international
grade standards to, meat animal carcasses or cuts, or for sorting for other
purposes (e.g. brand
names, product lines, etc.). The first step involves collecting image data
from the meat 10 using
the imaging device 12. The second step involves processing the image data
using the data
processing unit 16. The third step involves using the results of the
processing step in reporting
io quality-determining characteristics that can be used to augment USDA
graders in the assignment
of USDA Quality Grades, in reporting the areas of cross-sectional muscle
surfaces that can be
used to augment USDA graders in the assignment of USDA Yield Grades or other
international
grade standards, in reporting meat defect conditions, or in sorting the meat
10 based on specific
requirements of, for example, a brand-name or product line program. Using this
method, the
~ s grader or operator's limited time to analyze the meat 10 can be focused on
examining parameters
most readily examined by a person, providing the grader or operator with more
data for each
sample of meat 10 in the same amount of time, and allowing more accurate
prediction of
palatability and assignment of Quality Grade and Yield Grade than is currently
possible. Iri
addition, this method allows required computations to be performed more
quickly and accurately
zo than is currently possible.
The following example is included to demonstrate a preferred embodiment of the
invention. It should be appreciated by those of skill in the art that the
techniques disclosed in the
examples which follow represent techniques discovered by the inventor to
function well in the
practice of the invention, and thus can be considered to constitute preferred
modes for its
zs practice. However, those of skill in the art should, in light of the
present disclosure, appreciate
that many changes can be made in the specific embodiments which are disclosed
and still obtain
a like or similar result without departing from the spirit and scope of the
invention.
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
- 17-
Example I
Segregation of beef carcasses with very low probabilities of tenderness
problems
A population of 324 beef carcasses was examined in an effort to segregate a
subpopulation of carcasses with very low probabilities (< 0.0003) of having
ribeye shear force
s values of 4.5 kg or greater and subsequent unacceptably tough-eating cuts.
Of the 324 carcasses,
200 were certified to meet the above standard for tenderness.
Of the 324 head, 17 head were preselected for the tender subpopulation on the
basis of
expert-determined (beef scientist or USDA Grading Supervisor) marbling scores
of Modest,
Moderate, or Slightly Abundant, the three highest degrees of marbling in the
United States
io Standards for Grades of Carcass Beef.
In a second preselection step, 41 head of the remaining 307 were preselected
on the basis
of L*a*b* color. These carcasses exhibited a second principle component of
lean L*, a*, and b*
values of less than -0.70. Such low values of the combined variable have been
observed to
consistently indicate sufficient tenderness of the subsequent cooked lean.
is Third, 19 of the remaining 266 head were preselected on the basis of
marbling
distribution. Marbling distribution was determined, and the variance of
marbling distribution
was calculated, by an apparatus of the present invention. A variance of
marbling distribution of
less than 1.1 has been observed to consistently indicate sufficient tenderness
of the subsequent
cooked lean (i.e. a shear force value of less than 4.5 kg).
Zo In the final step, tenderness values for each of the remaining 247 head
were predicted
using a multiple regression equation using CIE a* values for lean and fat, as
well as machine
measured marbling percentage squared. The multiple regression equation
determined that 123
out of 247 carcasses were predicted to have a probability of being not tender
of 0.0003. These
123 carcasses were then segregated with the 77 that had been preselected, and
certified as being
Zs tender. The remaining carcasses had a normal probability of 0.117 of having
shear force values
in excess of 4.5 kg.
The results indicate that the system is able to segregate groups of beef
carcasses having
very low probabilities of unacceptable toughness, where current grading
methods cannot
perform this segregation adequately. This improved sortation increases the
economic value of
3o tender carcasses which would have been graded lower by current methods.
CA 02363089 2001-08-17
WO 00/49400 PCT/US99/19027
-18-
Both the apparatus and methods disclosed and claimed herein can be made and
executed
without undue experimentation in light of the present disclosure. While the
apparatus and
methods of this invention have been described in terms of preferred
embodiments, it will be
apparent to those of skill in the art that variations can be applied to the
apparatus and methods
s without departing from the scope of the present claims.