Research

The ability to measure quality of food products is a key parameter for the food industries
1) in sorting of raw materials;
2) in maintaining a uniform high quality product with minimum waste; and
3) in developing and producing food products with required chemical, physical, sensory, or health related properties.

The CIFQ project focuses on a new data driven design methodology that will result in a series of in-line and in process active illumination and computer vision based measurement systems. These non contact systems with zero contamination risk will capture and relate surface and sub-surface optical properties to food product quality. Primary cases are dairy and meat products.

Project plan

The project is organized in 5 work packages:

WP1
Develop, establish and maintain imaging instrument platforms including graphical user interface, workflow management, calibration, control of internal settings, database integration, image processing toolboxes, and speed optimization.
WP leader: Technical Director, Jens Michael Carstensen, Videometer A/S
Participants: NKT Photonics, DTU Compute

WP2
Physical modelling for simulation and inference of subsurface scattering including multiple types of scatterers and unmixing of scattering effects.
WP leader: Professor, Peter Andersen, DTU Fotonik
Participants: DTU Compute

WP3
Data driven vision system design including regression modelling and variable selection for establishing relatons between system configuration and food quality estimators.
WP leader: Associate Professor, Henrik Aanæs, DTU Compute
Participants: Danisco A/S

WP4
Dairy product quality assessment.
WP leader: Flemming Møller, Danisco A/S
Participants: Arla Foods Amba, DTU Compute, DTU Fotonik

WP5
Meat product quality assessment.
WP leader: Professor, Head of Section, Bjarne Ersbøll, DTU Compute
Participants: DMRI/Technological Institute, DTU Fotonik

PhD and postdoc projects

PhD1: Camilla Himmelstrup Trinderup "Spectral imaging of meat quality - color, texture, and structure"
Estimation of quality parameters on primarily beef cuts and half carcasses based on multispectral imaging systems including SLS.
Time: 2011/12 - 2014/12
Participants: DTU Compute, DMRI/Technological Institute
Supervisor: Professor, Knut Conradsen, DTU Compute

PhD2: Jacob Lercke Skytte - "SLS based dairy process control systems"
A primary case of study is fermentation of cultured milk products using SLS, but also whipped products processes will be considered.
Time: 2010/12 - 2014/03
Participants: Arla Foods Amba, Danisco A/S, DTU Compute, DTU Fotonik
Supervisor: Professor, Head of Section, Rasmus Larsen, DTU Compute

Postdoc3: "Quality inspection of dairy products"
Estimation of dairy product quality. The project focuses on prediction of content/mechanical properties/homogeneity of mousses/ice creams/cheeses using both SLS and diffuse light systems.
The postdoc position is devided between Postdoc3a: Jacob Lercke Skytte, time: 2014/03-2014/12 Postdoc3b: Otto Højager Attermann Abildgaard, time: 2014/03-2014/12
Participants: Arla Foods Amba, Danisco A/S, DTU Compute, DTU Fotonik

PhD4: Faisal Kamran - "Light propagation in highly turbid media"
Methods for solving light propagation in highly turbid media based on transport theory will be established. Then the aim is to establish the foundation for inverse methods, i.e., establishing the foundation for relating the observation(s) (measurement of scattered, diffuse light) to the underlying optical properties.
Time: 2011/01 - 2015/04
Participants: DTU Fotonik, DTU Compute, Danisco A/S
Supervisor: Professor, Peter Andersen, DTU Fotonik

Postdoc5: Anders Dahl - "The continuously reconfidurable vision workbench"
A reconfigurable vision system work bench will be constructed based on the DTU robot and relevant light sources. The system will be characterized and a new data driven paradigm for vision system design implemented.
Time: 2010/03 - 2012/02
Participants: DTU Compute, Videometer A/S, NKT Photonics

PhD6: Sara Sharifzadeh - "Multivariate analysis techniques for optimal vision system design"
Multivariate methods for regression/classification in highly structured data with an emphasis on obtaining sparse solution is researched.
Time: 2011/06 - 2015/05
Participants: DTU Compute, Danisco A/S
Supervisor: Professor, Head of Section, Bjarne Ersbøll, DTU Compute

PhD7: Otto Højager Attermann Abildgaard - "Optimization of super continuum light source for food inspection system"
In this project a a supercontinuum light source is optimized with respect to filters, spectrum, alternative configurations, diffuser, and polarization.
Time: 2010/06 - 2014/03
Participants: DTU Compute, Videometer A/S, NKT Photonics A/S
Supervisor: Associate Professor, Henrik Aanæs, DTU Compute