1. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
Focus set based reconstruction of micro-objects
Jan Wedekind
13.9.2004
MiCRoN http://wwwipr.ira.uka.de/~micron/
MMVL http://www.shu.ac.uk/mmvl/
People: Balasundram Amavasai, Manuel Boissenin, Axel B¨rkle,
u
Fabio Caparrelli, Arul Selvan, Jon Travis
microsystems & machine vision lab -1- MiCRoN http://wwwipr.ira.uka.de/~micron/

2. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
micro-vision and micro-robotics
Closed-loop control
• Estimate pose of manipulator in real time
Task planning
• Estimate pose of known micro-objects
• Recognize obstacles for avoiding collisions
• Provide data for determining gripping points
microsystems & machine vision lab -2- MiCRoN http://wwwipr.ira.uka.de/~micron/

3. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
present state of micro-vision
characteristics
• Limited depth of focus
• Teleoptical settings
vision problems
• unstable feature extraction
• limited depth of view
⇒ most standard approaches are failing
microsystems & machine vision lab -3- MiCRoN http://wwwipr.ira.uka.de/~micron/

4. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
spin images (Andrew Johnson 1997)
microsystems & machine vision lab -4- MiCRoN http://wwwipr.ira.uka.de/~micron/

5. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
spin images (Andrew Johnson 1997)
microsystems & machine vision lab -5- MiCRoN http://wwwipr.ira.uka.de/~micron/

6. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
depth of focus
• Acquire focus set of images g(x1 , x2 , z)
• Compute local sharpness measure s(x1 , x2 , z)
• Compute depth map d(x1 , x2 ) := argmax s(x1 , x2 , z)
z∈Z
microsystems & machine vision lab -6- MiCRoN http://wwwipr.ira.uka.de/~micron/

7. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
systematic error
test with micro-ridge
• Algorithm fails when contrast low
• Systematic errors
• Trade-off between resolution and stability
⇒ Filter-bench
image index z insufficient contrast PSF
d
c
b a
pixel−position in image y d c
systematic error
microsystems & machine vision lab -7- MiCRoN http://wwwipr.ira.uka.de/~micron/

8. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
multiscale approach
s maximum sharpness
• Recursive filtering
• Not real-time efficient
z
large filter kernel
• Complexity
s
(N = number of pixel)
– O(N ) time
– O(N ) memory z
√ 2 medium−sized filter kernel
– 3 N O(N 3 ) s
parallelisation
• No systematic error image
(x1,x2) z
small filter−kernel
microsystems & machine vision lab -8- MiCRoN http://wwwipr.ira.uka.de/~micron/

9. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
results
microsystems & machine vision lab -9- MiCRoN http://wwwipr.ira.uka.de/~micron/

10. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
results
microsystems & machine vision lab - 10 - MiCRoN http://wwwipr.ira.uka.de/~micron/

11. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
results
microsystems & machine vision lab - 11 - MiCRoN http://wwwipr.ira.uka.de/~micron/

12. Focus set based reconstruction of micro-objects- http://vision.eng.shu.ac.uk/mechrob04/MechRob-paper.pdf
results
microsystems & machine vision lab - 12 - MiCRoN http://wwwipr.ira.uka.de/~micron/