Multispectral imaging provides valuable information on the quality and safety of a vast array of materials from pharmaceuticals to raw meat and burned biscuits. The same techniques can be used to measure skin sensitivity to sticking plaster, detect counterfeit drugs and packaging and gain insight into historical artefacts such as medieval manuscripts and weapons. The VideometerLab 3 takes 19 images at 19 specific wavelengths in ultraviolet, visible and infrared light. With multivariate statistical analysis we can reveal hidden patterns in our 19-dimension colour space and find information that would be otherwise 'invisible' to our human eyes.

1. © analytikLtd
analytikLtd
Multispectral Imaging of Food Quality
Rapid Non-destructive Surface Analysis
Adrian Waltho, Analytik Ltd (Cambridge, UK)
adrian.waltho@analytik.co.uk
www.analytik.co.uk/multispectral-imaging
• Light reflectance at separate colours
• Statistical image processing
• Chemical, physical and spatial properties
• Differentiate | Quantify | Report

2. © analytikLtd
Traditional colour imaging uses
three broad bands of colour:
Red, Green and Blue
Normal Colour Imaging

3. © analytikLtd
Normal Colour Imaging
• RGB photographs have
limited spectral resolution
• Chlorophyll a and b give
almost the same RGB
signal and are not
spectrally separated
Chloro-a High Low High
Chloro-b High Low High

4. © analytikLtd
Multispectral Imaging
Chloro-a High Low Low High
Chloro-b Med High Med Low
• Using just 4 wavelength
bands with tightly defined
ranges, Chlorophyll a and b
can easily be distinguished
• VideometerLab 3 uses 19
wavelength bands

5. © analytikLtd
Multispectral Imaging
• Many images obtained at
selective wavelength bands
• Each image pixel contains
spectral data points
• Spectral signature reveals
chemo-specific information
• See spatial location of
surface chemical variation
Ultraviolet Near-Infrared
Infra-Red
Red
Yellow
Green
Blue
Ultra-Violet

6. © analytikLtd
VideometerLab 3 Schematic
• Narrowband illumination by 19
LEDs at points between 375nm-
970nm (UV-Vis-VNIR), strobing
sequentially with precise time
control for known light output
• Integrating sphere diffuses light
onto sample from all angles
• 5 mega-pixel monochrome CCD
camera captures % reflectance
at each LED wavelength for each
image pixel, NIST calibration
• Optional Emission filter wheel
for fluorescence detection and
Bright-field or Dark-field back
lighting
Camera
LEDs
Integrating
sphere
Emission
filter
wheel
Sample

7. © analytikLtd
VideometerLab 3 Schematic
• Precise control at each separate
point of the 375-970nm
spectrum, only possible with
LEDs, allows optimal illumination
for best signal to noise
• Provides calibrated data, suitable
for statistical analysis and
manipulation for discovery
• Emission filters allow additional
multispectral fluorescence
imaging, with excitation at each
LED wavelength
• Powerful, user-friendly hardware
setup and image analysis
software is central to the system
Camera
LEDs
Integrating
sphere
Emission
filter
wheel
Sample

8. © analytikLtd
VideometerLab 3 Schematic

9. © analytikLtd
Index of application areas
1. Sorting Categorising
1. Seafood sorting
2. Product composition
3. Pork fat percentage
2. Spoiling
1. Meat spoilage
2. Meat browning
3. Fruit oxidising
4. Biscuit burning
3. Quality Control
1. Seed contamination
2. Biscuit mishandling
3. Counterfeit packaging
www.analytik.co.uk

10. © analytikLtd
Sorting Seafood Sorting
Back to Index5
• The traditional method of assessing salmon colouration is to use a colour chart.
A human operator will offer his or her ‘best guess’ as to the colour of the salmon
This salmon is a 27 on the
SalmoFan chart opposite.

11. © analytikLtd
Sorting Seafood Sorting
Back to Index5
• The VideometerLab can be taught two extremes of salmon colouration, and
then will false colour an image to represent what it has just learned (blue is the
‘pinkest’ a salmon can ever be, orange is the ‘reddest’)

12. © analytikLtd
Sorting Seafood Sorting
Back to Index5
• It can then apply this model to a real salmon, and will not make any of the
mistakes the traditional methods will
Colour in RGB False Colour Image
(Human Vision) (Videometer ‘Vision’)

13. © analytikLtd
Sorting Seafood Sorting
Back to Index5
• Most usefully, a salmon that does not fall neatly onto an integer number can
have its colour determined with up to two decimal places of accuracy

14. © analytikLtd
Sorting Seafood Sorting
Back to Index5
• Prawn sorting is done in the same way, with the same sorts of subjectivity
problems, compounded by the fact there are many hundreds of prawns to be
analysed at any one time

15. © analytikLtd
Sorting Seafood Sorting
Back to Index5
• The VideometerLab can separate touching prawns and automatically put them
in the correct ‘bin’ to guide an operator later. In this image the VideometerLab
has automatically ignored anything too small to be a prawn

16. © analytikLtd
Sorting Product Composition
Back to Index5
• In this image the operator wants to identify what percentage of this tuna steak is
blood meat

17. © analytikLtd
Sorting Product Composition
Back to Index5
• The operator can either highlight the difference between blood meat on the left
and other meat on the right and measure it with the onboard geometry tools or
skip out the human element altogether and quantify the percentage
automatically, as in the right-hand image

18. © analytikLtd
Sorting Product Composition
Back to Index5
• Similar techniques can be used on most heterogeneous substances, such as
the morceau in this foix gras

19. © analytikLtd
Sorting Pork Fat Percentage
Back to Index
• The VideometerLab can quantify the percentage of fat in this
minced pork

20. © analytikLtd
Sorting Pork Fat Percentage
Back to Index
• It does this by learning the spectral signature of fat and pork,
then correlating the two

21. © analytikLtd
Sorting Pork Fat Percentage
Back to Index
• In this experiment the VideometerLab performs excellently –
it is always within 0.5% of the chemical test’s predictions
VideometerLab%Estimate
Chemical Analysis % Fat

22. © analytikLtd
Spoiling Meat Spoilage
Back to Index
• The VideometerLab can be used for a variety of meat
spoilage tests – here is one example looking for the number
of bacterial colonies and discolouration of some mince after
storage under different conditions

23. © analytikLtd
Spoiling Meat Spoilage
Back to Index
• These images show just the UV (405nm) band to better
highlight the size and distribution of the bacterial colonies
No special storage Atmospheric Packaging M.A.P.

24. © analytikLtd
Spoiling Meat Spoilage
Back to Index
• This histogram demonstrates quantitatively what is visually
apparent on the last slide – MAP reduces the incidence and
size of bacterial colonies
• Sample 1 before storage
• Sample 1 after atmospheric
storage
• Sample 2 before storage
• Sample 2 after MAP storage

25. © analytikLtd
Spoiling Meat Discolouration
Back to Index
• Alternatively, it can be used to predict browning, as with this
sample of steak – the blue area shows higher levels of
metmyoglobin which appears to predict discolouration

26. © analytikLtd
Spoiling Fruit Oxidising
Back to Index
• The VideometerLab can be used as an objective measure of
food spoilage

27. © analytikLtd
Spoiling Fruit Oxidising
Back to Index
• The blue areas are browning, the red areas are unaffected. This can be
quantified automatically by the VideometerLab software

28. © analytikLtd
Spoiling Fruit Oxidising
Back to Index
• The VideometerLab can capture the whole browning process
1 5 10
15 20

29. © analytikLtd
Spoiling Biscuit Burning
Back to Index
• In this slide the operator has taken a spectrum of the corner
of each of these biscuits, and proved that the ‘blue’ biscuit
(top left) has been burned during cooking

30. © analytikLtd
QC Seed Contamination
Back to Index
• A human operator would be hard pushed to notice the single poppy seed
contaminating this dish of sesame seeds. This could be fatal if instead of poppy
seed it were a fragment of peanut, for example

31. © analytikLtd
QC Seed Contamination
Back to Index
• Magnification of the previous image
1mm

32. © analytikLtd
QC Biscuit Mishandling
Back to Index
• An operator handling this biscuit with wet fingers will spoil
the taste, but may be difficult to spot if it is unexpected

33. © analytikLtd
Multispectral Image Analysis
• A variety of image processing tools
are available in VideometerLab’s
analysis software
• Analysis protocols and models can
be saved and applied to new
samples for routine analysis
• Multiple analysis models can be
run on the same image data to
quantify separate parameters
• Analysis can be batch-run to aid in
high-throughput data acquisition
• Image archiving provides a record
of samples for future data mining
and regulatory compliance

34. © analytikLtd
• Nuanced, human-like analysis of
heterogeneous samples
• Objective, accurate, repeatable
measurements
• Non-destructive analysis allows
further sample assay work
• No sample preparation is required
• Rapid data acquisition and analysis
enables high throughput screening
Why use Multispectral Imaging?

35. © analytikLtd
• Hardware options
• Emission filter wheel (4 x 25mm Ø) for
fluorescence macro-imaging
• Dark- or Bright-field backlighting for scattering
or translucency analysis
• Bespoke solutions available
• Software options
• Image Processing Toolbox for filtering and
segmentation of image data in individual
spectral bands
• Multispectral Imaging Toolbox for multivariate
image analysis and modelling
• Blob Toolbox for granular object separation,
analysis and data output reporting
• Matlab multispectral imaging toolbox integration
Hardware and Software Options

36. © analytikLtd
• Nuanced, human-like analysis of
heterogeneous samples
• Sophisticated data analysis tools
for the novice and power user
• Objective, repeatable data
collection with easy export
• Rapid, non-destructive testing with
no sample preparation
VL3 Multispectral Imaging
adrian.waltho@analytik.co.uk
Multispectral imaging at Analytik.co.uk
VL3 Datasheet PDF