Development of broadband multispectral imaging
for analysis of dairy products
Ph.D. student Otto H. A. Nielsen
Technical University of Denmark, department of InformaticsCentre for imaging food quality (CIFQ)
VideoMeterNKT - photonics
DANISCO
Photonex 2010 conference on spectral imaging3th November 2010
DTU - Informatics 1/18
Outline
1 SLS based hyperspectral imaging
2 Illumination system
3 Preliminary results
4 Outlook
DTU - Informatics 2/18
Outline
1 SLS based hyperspectral imaging
2 Illumination system
3 Preliminary results
4 Outlook
DTU - Informatics 3/18
Spectral imaging for food diagnostics
DTU - Informatics 4/18
Spectral imaging for food diagnostics
Vision systems
Structural sampleinformation
Characterisation ofbiological samples
Spectroscopy
Chemical analysis
DTU - Informatics 4/18
Spectral imaging for food diagnostics
Vision systems
Structural sampleinformation
Characterisation ofbiological samples
Spectroscopy
Chemical analysis
Advantages for food analysis:
Fast
Non-invasive
⇒ Suitable for in line process inspection
DTU - Informatics 4/18
Inspection method for homogeneous products milk,
cream, yoghurt etc.
Illumination sources, a collimated beamSub surface light scattering (SLS)
DTU - Informatics 5/18
Inspection method for homogeneous products milk,
cream, yoghurt etc.
Illumination sources, a collimated beamSub surface light scattering (SLS)
Scattering of biologicalparticles is described myLorentz-Mie theory.
DTU - Informatics 5/18
Inspection method for homogeneous products milk,
cream, yoghurt etc.
Illumination sources, a collimated beamSub surface light scattering (SLS)
Scattering of biologicalparticles is described myLorentz-Mie theory.
Can determine:
Particle size
Particle concentration
Light absorption
Light scattering
DTU - Informatics 5/18
Previous measurements performed by Danisco1.
1On-line monitoring of food processes using subsurface laser scatteringDTU - Informatics 6/18
Previous measurements performed by Danisco1.
1On-line monitoring of food processes using subsurface laser scatteringDTU - Informatics 6/18
Previous measurements performed by Danisco1.
The structure of the sample relates it’s optical properties. Thecharacterisation method will be described later.
1On-line monitoring of food processes using subsurface laser scatteringDTU - Informatics 6/18
Outline
1 SLS based hyperspectral imaging
2 Illumination system
3 Preliminary results
4 Outlook
DTU - Informatics 7/18
Illumination source
Light generation with a broad spectral profile
SuperK light source
Produces a bright output from 500 nm to 2500 nm, and the beamrequires optical filtering to be used in spectral imaging.
DTU - Informatics 8/18
Optical filtering ⇒ hyperspectral beam
Acousto-optical tunable filter (AOTF)
Can cover the spectrum from500 nm to 2500 nm with twoAOTF crystals
Returns two beams, one with visible light and a one for the NIR.
DTU - Informatics 9/18
Optical filtering ⇒ hyperspectral beam
Acousto-optical tunable filter (AOTF)
Can cover the spectrum from500 nm to 2500 nm with twoAOTF crystals
Returns two beams, one with visible light and a one for the NIR.
DTU - Informatics 9/18
Final light delivery
The light is delivered using an micro structured fiber (LMA5).
500 550 600 650 700 750 800 850 9000
100
200
300
400
500
600
Wavelength [nm]
Bea
m w
aist
[µ
m]
500 550 600 650 700 750 800 850 9000
0.1
0.2
0.3
0.4
0.5
Wavelength [nm]
beam
pow
er [
mW
]
DTU - Informatics 10/18
Schematically setup
DTU - Informatics 11/18
Schematically setup
Novel features of the illuminations system
An small fibre output → easy to handle
Collimated Gaussian beam → simple scattering pattern
Broad spectral range (500 nm to 900 nm) → chemicalspecificity
Fast switching of wavelengths → usable for industrialinspection
DTU - Informatics 11/18
Outline
1 SLS based hyperspectral imaging
2 Illumination system
3 Preliminary results
4 Outlook
DTU - Informatics 12/18
Image example with whole milk, λ = 670 nm
DTU - Informatics 13/18
Image example with whole milk, λ = 670 nm
DTU - Informatics 13/18
Image example with whole milk, λ = 670 nm
The scattering can be characterised by the slope of the loglogcurve and the spot size of the saturated.
DTU - Informatics 13/18
SLS slope characteristics of diary products
500 600 700 800 9000
0.5
1
1.5
2
2.5
3x 10
−3
SLS
Slop
[ar
b. u
nit]
Wavelenght [nm]
Skimmed milk 0.5 %letmælk 1.5%Whole milk 3%
Can discriminate betweenproducts of different fat contents.
DTU - Informatics 14/18
SLS slope characteristics of diary products
500 600 700 800 9000
0.5
1
1.5
2
2.5
3x 10
−3
SLS
Slop
[ar
b. u
nit]
Wavelenght [nm]
Skimmed milk 0.5 %letmælk 1.5%Whole milk 3%
Can discriminate betweenproducts of different fat contents.
500 550 600 650 700 750 800 8500
0.5
1
1.5
2
2.5
3x 10
−3
SLS
Slop
[ar
b. u
nit]
Wavelenght [nm]
yoghurt 3 % fatCream 9 % fatCream 18 % dilutedCream 18 % fat
Different slope for same fatcontents
DTU - Informatics 14/18
Spot size analysis of diary products
500 550 600 650 700 750 800 8500
100
200
300
400
500
600
Spot
siz
e [p
ixel
s]
Wavelenght [nm]
Skimmed milk 0.5 %letmælk 1.5%Whole milk 3%
DTU - Informatics 15/18
Spot size analysis of diary products
500 550 600 650 700 750 800 8500
100
200
300
400
500
600
Spot
siz
e [p
ixel
s]
Wavelenght [nm]
Skimmed milk 0.5 %letmælk 1.5%Whole milk 3%
500 550 600 650 700 750 800 8500
50
100
150
200
250
300
350
400
450
Spot
siz
e [p
ixel
s]Wavelenght [nm]
yoghurt 3 % fatCream 9 % fatCream 18 % dilutedCream 18 % fat
DTU - Informatics 15/18
Spot size analysis of diary products
500 550 600 650 700 750 800 8500
100
200
300
400
500
600
Spot
siz
e [p
ixel
s]
Wavelenght [nm]
Skimmed milk 0.5 %letmælk 1.5%Whole milk 3%
500 550 600 650 700 750 800 8500
50
100
150
200
250
300
350
400
450
Spot
siz
e [p
ixel
s]Wavelenght [nm]
yoghurt 3 % fatCream 9 % fatCream 18 % dilutedCream 18 % fat
Line crossings at long wave lengths, showing interesting features.
DTU - Informatics 15/18
Outline
1 SLS based hyperspectral imaging
2 Illumination system
3 Preliminary results
4 Outlook
DTU - Informatics 16/18
Future implementations
Analyse products with simple changes (fat contents andparticle size). Make correlation between the SLS spectrumand the food property based on statistical models.
Combining the illumination system with a robot to move thefibre. This will make it possible to examine non symmetricscattering distribution and surface reflection.
Perform instrument independent characterisation of thesamples intrinsic scattering parameters.
DTU - Informatics 17/18
Acknowledgements
The work presented is sponsored by the The Danish Council forStrategic Research
Coworkers:
Anders Dahl - post.doc. DTU IMM - projectguidance and data analysis
Videometer - imaging equipment and dataanalysis
NKT photonics - construction and tuning of theillumination system
Danisco - introduction to food analysis
DTU - Informatics 18/18