Measurement of food colour and structure using digital...
Transcript of Measurement of food colour and structure using digital...
Measurement of food colour and
structure using digital imaging
Martin Whitworth
Campden & Chorleywood
Food Research Association
www.campden.co.uk
Contents
• Colour imaging and measurement of
foods
• C-Cell analysis of bread structure
• X-ray computed tomography of
structural changes during baking
• Hyperspectral near infrared imaging of
food composition
Colour imaging
Assessment of food colour
• Colorimeters are used for quality
control, including on-line
measurement.
• Coloured reference tiles and cards
are used as references for some
specific applications.
• Photographs and printed images
are widely used as visual
references in production areas;
several retailers require suppliers
to provide product images.
Colour imaging
• Digital images and prints
are widely used for
documentation,
specification and quality
control.
• Colour reproduction is
often poor and
inconsistent.
• Calibrated imaging systems can provide better
specification images.
• Image analysis can be used to measure aspects such
as colour distribution.
Colour imaging facilities at CCFRA
DigiEye
imaging
system
Profiled
monitorD65
viewing
cabinet
Monitor
profiling
device
Eye-One iO
spectrophotometer
Profiled
printer
Colour management
software:
DigiEye
Photoshop
Profile Maker
CCFRA software
Image capture• DigiEye imaging system
• Controlled illumination conditions
• Calibration against CIELAB colour space
Camera calibration
• Set white balance.
• Take image of uniform
white card to correct for
spatial variations in
illumination.
• Take image of test card
with 240 known
colours.
• Calculate mapping
between reference XYZ
colour values and
measured RGB values.
Printing
• Epson Stylus Photo R2400
printer
• Lyson continuous ink
system used to reduce ink
costs
• Calibrated using a custom
ICC profile
Creating an ICC printer profile
Printer driver
RGB data
Ink loading
CIELAB data
Printer profileRGB test image
Measure print
Calculate profile
specifying which
RGB values to
send to printer to
achieve the
required CIELAB
colour
Printed image
Editing profile for use with DigiEye
Edit profile to
match measured
colours to original
CIELAB data.
Printer driver
RGB data
Ink loading
CIELAB data
Printer profile
Printed image
Measure print
with DigiEye
CIELAB test image
Printer profiling performance
Mean ΔE=2.2 for DC Colorchecker chart
Products Profiled print
Development of colour scales
• Aim to develop a systematic approach for
producing sets of images for food assessment
against defined colour scales.
• Approach:
• Identify relevant attributes by sensory profiling
• Develop image analysis measurements of these
attributes and define a numerical scale.
• Determine the spacing of distinct points on the scale
from the sensory data.
• Select images at these points to produce a chart.
Case study: Bread roll crust colour
• Rolls baked with 36
recipe and process
variations
• Measurements:
• Sensory descriptive
analysis by 9 assessors
• Colorimeter
• DigiEye images
• Diffuse
• Diffuse + oblique
a*
b*
L*
Degree of bake, λLight
Dark
Roll crust colour measurements(DigiEye images, diffuse, 10mm central region)
• Scale assigned
based on fitting
a principal curve
• Reduces colour
variation to a
single parameter
• Provides a
general
approach for
defining colour
scales
Distribution of degree of bake
30
40
50
60
70
80
90
100
20
10
0
Degree
of bake
30
35
90
8090
30
20
65
70
7060
70
252590
Synthetic images
• Colour of each pixel shifted parallel to the curve.
• More realistic effect than a global colour shift.
• Could be used to specify acceptable colour
limits.
λ=30 λ=50 λ=70
Image analysis
Cherry cake
a*
b*Crumb Cherries
Ruler
Background
1330 (10.2%)
1409 (11.1%)
Area of
cherries (mm2)
Crumb colour
28.9 3.413.4 5.475.7 6.91808213061Bottom
29.4 3.914.8 5.773.5 8.91798112714Top
b*a*L*
Width
(mm)
Height
(mm)
Area
(mm2)
Crust colour variation
0
10
20
30
40
50
60
70
80
90
0 2 4 6 8 10 12 14
Distance from surface (mm)
Colo
ur
valu
e
Top crustLeft crustRight crustBottom crust
L*
b*
a*
Distribution of cherries
0
200
400
600
800
1000
1200
1400
1600
0 10 20 30 40 50 60 70 80
Height within slice (mm)
Are
a o
f ch
err
ies (
mm
2)
Upper slice based on discrete cherries
Lower slice based on discrete cherries
Upper slice based on integrated area
Lower slice based on integrated area
0
100
200
300
400
500
600
700
800
0.0 - 2.5 2.5 - 5.0 5.0 - 7.5 7.5 - 10.0 10.0 - 12.5 12.5 - 15.0 15.0 - 17.5 17.5 - 20.0
Diameter range (mm)
Are
a o
f ch
err
ies (
mm
2)
Upper slice
Lower slice
Size
distribution
Distribution
with height
C-Cell image analysis system
• Designed for bakery use.
• Sliced products imaged
under oblique illumination,
providing good contrast of
structure.
www.c-cell.info
• Analysed to measure
structure and dimensions.
• 50 instruments in use in
15 countries.
Measurement of cell sizes in bread
davg=2.00mmdavg=1.27mm
davg=2.20mmdavg=1.61mm
Cell
are
a (
% o
f sli
ce a
rea)
Cell diameter (mm)
Cell
are
a (
% o
f sli
ce a
rea)
Cell diameter (mm)
Cell
are
a (
% o
f sli
ce a
rea)
Cell diameter (mm)
Cell
are
a (
% o
f sli
ce a
rea)
Cell diameter (mm)
High magnification for extrudates
250
0 0.5 1.0 1.5
300
350
CaCO3 (% of maize mass)
Scre
w s
peed
(rp
m)
Cell diameter (mm)
1.82
1.35
1.00
1.86
1.44
1.16
1.89
1.29
1.12
1.90
1.31
1.10
(Mean of 3
samples)
Area (cm2)
1.59
1.45
1.16
1.48
1.26
1.21
1.47
1.27
1.16
1.61
1.25
1.09
X-ray computed tomography
imaging of bakery processes
X-ray computed tomography of baking
• Products are baked in an
insulated glass chamber.
Heater
Fan
Baking chamber
CT scanner• Baking is carried out using
hot air blown through the
chamber.
• Non-destructive images are taken during processing
Final proof
• Single piece moulding
creates a spiral
structure in the dough.
• During proof, the
dough expands to fill
the pan and then
expands vertically.
Bread baking• A break forms in the crust
early during baking.
• Internal expansion causes oven spring.
• A zone of expansion moves inwards.
• Peripheral structures are compressed against the pan walls.
• As the crust becomes firm,
oven spring ceases and
material is compressed
against the underside of the
crust.
Baking processes in cakes
Sponge
MuffinsHigh ratio cake
1.0
0.1
0.0
0.2
0.3
0.4
0.5
0.6
0.7
0.9
0.8
Relative
absorbance
Air
Water
Hyperspectral near infrared imaging
of food composition
Hyperspectral near infrared imaging• Gilden Photonics/Specim
instrument
• Takes images with a full NIR
spectrum for each pixel.
• Cooled HgCdTe camera
sensitive over 900-2500nm
NIR wavelengths, fitted with
spectrograph.
• Field of view ~8mm to 30cm.
• Scan time ~5s
• Enables mapping of food
composition.
Method
• Data cube collected
• Sliced for analysis
Width of deck
Wavelength
900nm
2500nm
Scan
direction
Spectra for each pixel
Images for each wavelength
Mean spectra for selected regions
Doughnut
• Measure areas under peaks and plot distribution
• Doughnut loses water and absorbs fat at the
surface during frying
Fat Water
Baguette moisture
Unbaked
1 hour 5 hours 24 hours 48 hours 96 hours
Stored at 65% R.H.
Stored at 80% R.H.
Mois
ture
(%
)
10
20
30
40
50
60
45.4 45.6 44.3 40.5 26.4
45.7 45.4 44.7 36.2 18.0
43.1
Plots: NIR data
Values: Reference values for central 20mm discs
Conclusions
• Calibrated colour imaging can improve the accuracy of
images used for food specification. Image analysis can
provide new quantitative measurements.
• The C-Cell imaging system provides practical objective
measurements of baked product structure.
• X-ray CT imaging enables the formation of these
structures to be studied with applications for product
development, fault diagnosis and training.
• Hyperspectral NIR imaging is a new technology for
mapping food composition. Applications include
measurement of fat and moisture migration.