quality improvement due to hyperspectral...
TRANSCRIPT
© Fraunhofer IOSB 1
Quality improvement due to
hyperspectral analysis
Henning Schulte Jena, 28.08.2014
Fraunhofer-Institut für Optronik, Systemtechnik und
Bildauswertung IOSB
Karlsruhe Ettlingen Ilmenau Lemgo
© Fraunhofer IOSB 2
Speech 28.08.2014 / H. Schulte
Partner for Sorting Technology
Spectral Signature – How to Use
Project Example => GrapeSort
Call für Papers => OCM-2015
Partner for Sorting Technology
Spectral Signature – How to Use
Project Example => GrapeSort
Call für Papers => OCM-2015
© Fraunhofer IOSB 3
Long Time Sorting Know HowFood, Glas, Minerals, Plastic, …
Busines Unit - Automated Visual Inspection
© Fraunhofer IOSB 4
Ore, Minerals
Plastic Granules
Glassrecycling
Grain, Corn, Coffe
Tea, Herbs
References and Products - Sorting
© Fraunhofer IOSB 5
Systems for bulk good sorting
© Fraunhofer IOSB 6
1400 nm 2100 nm
E. Neil Lewis u. a., „Near‐infrared Spectral Imaging with Focal Plane Array Detectors“ (o. J.): 25-55.
Hyper-Spectral-Analysis
Laboratory equipment(240 – 2500 nm)
© Fraunhofer IOSB 7
Speech 28.08.2014 / H. Schulte
Partner for Sorting Technology
Spectral Signature – How to Use
Project Example => GrapeSort
Call für Papers => OCM-2015
Spectral Signature – How to Use
© Fraunhofer IOSB 8
Infrastructure Spectral Data
Multispectral
Measuring
Systems240nm – 2.500nm
Web-
BrowserSQL-Database
Further development of
evaluation methods with
MATLAB etc.
Standard-Analysis-
Tools for customers
Online
Analysis
Multispectral
Measuring
Systems240nm – 2.500nm
Further development of
evaluation methods with
MATLAB etc.
Online Analysis
© Fraunhofer IOSB 9
Speech 28.08.2014 / H. Schulte
Partner for Sorting Technology
Spectral Signature – How to Use
Project Example => GrapeSort
Call für Papers => OCM-2015
Project Example => GrapeSort
© Fraunhofer IOSB 10
Quality Improvement of Wine – GrapeSort
Research project
INWaG
Ingenieurbüro Waidelich
Feb. 2013 – Jan. 2015
© Fraunhofer IOSB 11
Quality Improvement of Wine – Current Situation
Constraints typical for winemaker
Harvest ~ 6 – 10 weeks
Material changes continuously
Color
Rigidity
Quality
Reduced qualified manpower
Resulting challenges
Easy to handle
Simple adaptation
Applicable for different grapes
Sorting to achieve various
compositions
Source: Fraunhofer IOSB
Material changes
continuously
Simple and for
different grapes
© Fraunhofer IOSB 12
Quality Improvement of Wine – Classification
Low sweetness
High sweetness
Source: Fraunhofer IOSB
Experiments
© Fraunhofer IOSB 13
Quality Improvement of Wine – Some Spectra
Pinot
Noir
Source: Fraunhofer IOSB
Source: Fraunhofer IOSB
Pinot
Blanc
© Fraunhofer IOSB 14
Quality Improvement of Wine – Transfer Data -> HW
Source: Fraunhofer IOSB
Data acquired in Oct. 2013 optical filter for prototype
in 2014
Data acquired in Aug./Sep.
2013
optical filter for
experimental sorter in
2013
© Fraunhofer IOSB 15
Quality Improvement of Wine – Classification Results
Pixel classification results for filter simulation
Hyperspectral
LDA
RGB + 3 filters
(Genetic alg.)
RGB + 3 filters
(Greedy alg.)
RGB + 1 filter
(Brute-force)
Pinot NoirSWIR 79% 62% 70% 60%
VIS/NIR 85% 74% 74% 74%
Pinot BlancSWIR 71% 62% 64% 62%
VIS/NIR 80% 68% 68% 66%
RieslingSWIR 81% 67% 74% 53%
VIS/NIR 85% 64% 63% 57%
© Fraunhofer IOSB 16
Source: Fraunhofer IOSB
Quality Improvement of Wine – Summary
Request for grape sorters increases
Hyperspectral data better than just additional
bandpass filters
Difference in mean value of Oe achievable with one
additional VIS/NIR-filter
17° Oe for Pinot Noir
9° Oe for Pinot Blanc
6.5° Oe for Riesling
© Fraunhofer IOSB 17
Speech 28.08.2014 / H. Schulte
Partner for Sorting Technology
Spectral Signature – How to Use
Project Example => GrapeSort
Call für Papers => OCM-2015 Call für Papers => OCM-2015
© Fraunhofer IOSB 18
CALL FOR PAPERSSubmission of Abstracts:
September 22nd , 2014
OPTICAL CHARACTERIZATION OF MATERIALSInternational Conference
MARCH 18-19, 2015 Karlsruhe, Germany
WWW.OCM-2015.EU
OCM
2015
M A R C H 18-19 , 2015
KARLSRUHE // GERMANY