Visions%20of%20The%20Virtual%20Slaughterhouse

1
Visions of The Virtual Slaughterhouse

• Søren G. Erbou
• Ph.D. student
• Informatics and Mathematical Modelling
• Technical University of Denmark
• Danish Meat Research Institute
• sge_at_imm.dtu.dk

2
Outline

• What is the The Virtual Slaughterhouse?
• Where are we now?
• Visions for the future
• Means
• Applications
• Summary

3
The Virtual Slaughterhouse

• Part of the Danish Meat Research Institute
  (Roskilde, 1954)
• Owned by the Danish pig producers via the Danish
  Meat Association
• Mission of DMRI
• Leading knowledge centre within meat and
  slaughter technology
• The Virtual Slaughterhouse (2006-08)
• 3D-models of pig carcasses and tools for
  analysing the models
• Collaboration with IMM, University of Århus and
  Visiana
• Own CT-scanner
• 4 phd-students
• Why?

4
Where are we now?

• Pig slaughter automation programme (1998-gt)
• gt30 development projects, gt40 mill.
• Improved working environment, hygiene and safety
• Examples

Cleaning of throat- and heart regions
Removal of inner bones from front part
5
Where are we now?

• Highly automated production
• Need for better quality control
• Reverse problem in the production
• Normal Many small parts assembled into one final
  product
• Reverse Same input disassembled into many
  products
• Large variability of input
• Products restrict each other
• Minimise variability on output and maximise
  profit
• New methods for production planning
• Operations research
• Define new input features

6
Visions for the future

• Where do we want to be in 10 years?
• More adaptable production
• Better suited for niche production
• Less variability in quality of the final products
• Handle large variability of input
• Keep the leading technological position worldwide

7
Means

• Modelling the biological variability
• Modelling specific cuttings
• New predictors of quality based on models
• Optimise the cutting of each carcass separately
• Introducing the CT-technology to the
  slaughterhouse industry
• Image analysis is the key for extracting only the
  necessary information

8
Online CT

• Adapt cutting to each specific carcass
• Unlimited amount of information available at an
  early stage along the slaughterline
• Extract useful information
• Define predictors for quality
• Speed and cost is crucial
• Minimise data acquisition while maximising useful
  information

9
Applications

• Trimming of the fat layer on the loin or neck
  muscle
• The same fat thickness all over the loin gives
  the best price
• Small thickness is better
• Important not to expose the muscle
• CT can give additional information
• Model the fat layer
• Fit the model to new carcasses

(University of Nebraska, Lincoln)
10
Applications

• Use models for developing robotic tools
• 40 carcasses CT-scanned
• 3D statistical shape model of bone structures in
  the ham.
• 7 model parameters describe 69 of the variation
• Parameters are localised to ease interpretation

11
Even more applications

• Segmentation of muscles
• Cutting and quality estimation of pig backs
• Pig atlas
• Volume registration
• Apply cuts in atlas and propagate to population
• Virtual dissection
• Voxelwise classification
• Better quality measure
• Many more to come

12
Summary

• Obtaining and analysing 3D-models of pig
  carcasses
• Introducing CT in the slaughterhouses
• Image analysis and statistics are key elements
• Predictors for quality
• From models to robotic tools
• Better suited for niche products
• Production more adaptable
• Quality control is crucial

13
Aknowledgements

• DMRI
• Eli V. Olsen
• Lars B. Christensen
• Claus Borggård
• IMM-DTU
• Martin Vester-Christensen
• Mads F. Hansen
• Peter S. Jørgensen
• Allan Lyckegaard
• Rasmus Larsen
• Bjarne K. Ersbøll,
• Thank you for your attention