wool testing grant · 2016. 2. 10. · wool testing grant christopher schauer ... reid redden,...
TRANSCRIPT
-
Wool Testing Grant
Christopher Schauer, Hettinger Research Extension Center
Igathinathane Cannayen, Ag & Biosystems Engineering
Reid Redden, Animal Science Department*
(Currently with Texas A&M)
-
Background
• ASI convention 2012 asked if a cell phone
could take a picture and “Grade” Wool
– Willing to sacrifice some accuracy to have an
inexpensive, widely available software to
analyze wool.
• Establish Collaborative Scientist
– Igathi Cannayen, Ag & Biosystems
-
Background
• Cell Phone Camera – not good enough
• Scanner (19200 DPI – 1.3 micron
resolution) and software analysis
-
Plan of Work
Possible system modification leading to
simpler layout (similar to laser instrument)
– but intelligent program evaluating from fiber
clusters of simpler layout
Time consuming layout Simpler layout
-
Budget
• Transportation – $4,500
• Labor - $8,500
• Equipment - $6,000
• Publication - $1,000
• Indirects - $2,000
• Total - $22,000
– Delivery of 3 “units” with the software installed
-
Report of Tasks Completed
-
• Single fiber layout – plugin developed
• Steel wool fibers – measured
• Effect of weight in fiber layout – studied
• Calibration using microscope – completed
• Tungsten wire calibration – completed
• Running same samples – OFDA & Scanning – calibration
• Development of glass cover & SS plate
• Development of feature rich plugin – OFDA calibration
Delivery of NDSU OonSA (Wool Scan Analyzer and
user manual) today on a USB key!
Development path and update from last year:
-
Snapshot of overall measurement procedure
using NDSU OonSA
-
• Handling multiple fiber layout
• 19,200 DPI = 1.3 mic; also handle higher DPI images
• Simpler interface; DPI, Wool type, Ear tag#
• Inbuilt OFDA based calibration
• Results in 2 textual and 2 graphical formats
• Several cutoff factors to select less variation chosen
• Plugin integrated in Fiji/ImageJ
• Optionally operated by shortcut
• Windows and Mac versions
• Entire software (Fiji + OonSA) comes in a flash drive
• Future updates can be communicated through *.jar files
Features of NDSU OonSA
-
Details of Procedure – Scanning Wool
• Scan preview showing the scanner bed (black), transparent sheet, glass cover, stainless steel frame, and wool sample
• Advanced mode for selecting 0.5” by 0.5” scan window and 19200 DPI
-
Fiji/ImageJ and NDSU OonSA
• Fiji/ImageJ is the basic open source image processing system• NDSU OonSA is the developed plugin – installed in Fiji• Windows and Mac version already developed
Details of Procedure
-
Opening Image & Plugin Operation
• Image can be opened by drag & drop or by File > Open• Plugin input panel receives simple inputs
Details of Procedure
-
Multiple fiber layout
• Measures shape factors (area, roundness, solidity, and aspect ratio)• Eliminates the overlap (e.g., segment 5) based on cutoff values
Details of Procedure
-
Textual output - Log
• Results summary in the form of log window • Accumulates results
Details of Procedure
-
Graphical outputs – Frequency & Widths
• Fifty frequency intervals – width vs % of fibers• accumulates results
Details of Procedure
-
Interpretation of results
• Red numbers - labeled measurements used in the analysis• Blue crosses indicate no overlap segments - measured but not included in
the analysis – due to excess variation and touching grids• Gray lines indicate 12 actual width measurements – average used• Black lines grids for digitally chopping the fibers
Details of Procedure
-
Spreadsheet – Storing results
• Spreadsheet output of the results analysis summary – automatic generation• New results are appended – whole history of results stored • Contents copied for storage/report or further analysis – friendly format
Details of Procedure
-
Results
• Set of washed wool and greasy wool tested
• Same samples subjected to OFDA and OonSA plugin
• Calibration equations using (i) U.S. grade diameter chart
and (ii) OFDA measurements were developed
• Both calibration equations were comparable
• OFDA calibration equations were used in OonSA plugin
• Running more samples – different sets will improve the
calibration equation – hence accuracy of OonSA
-
US grade chart calibration
• Calibration model (rather than single factor) was used• Washed wool lines up well • Greasy wool older samples – reduced R2 - same US chart values used – no
correction for grease applied
Results
y=1.1899x- 40.064
R²=0.90748
0
5
10
15
20
25
30
35
40
45
40 45 50 55 60 65 70
U.S.WoolStd.Dia(micron)
OonSAdirectvalues(micron)
Washedwool
y=1.2884x- 46.404
R²=0.721910
5
10
15
20
25
30
35
40
45
40 45 50 55 60 65 70
U.S.W
oolStd.Dia(micron)
OonSAdirectvalues(micron)
Greasywoll
-
US grade chart calibration – values compared
• Variation exists even between US chart value and OFDA (Blue & Gray bars)• Washed wool trends are better • Greasy wool samples more variation among methods - older samples - same
US chart values used – no correction for grease applied
Results
0
5
10
15
20
25
30
35
40
45
36 44 46 48 50 54 56 58 60 62 64 70 80
Woolfiberdiameter(m
icron)
Grade
Washedwool
Actual OFDA OonSA
0
5
10
15
20
25
30
35
40
45
36 40 44 46 48 50 54 56 58 60 62 64 70 80
Woolfiberdiameter(m
icron)
Grade
GreasywoolActual OFDA OonSA
-
OFDA & OonSA calibration
• Calibration equations were similar for US grade values and OonSA• These equations used in OonSA plugin
Results
y=1.1851x- 39.92
R²=0.89404
0
5
10
15
20
25
30
35
40
45
40 45 50 55 60 65 70
OFDADia(m
icron)
OonSAdirectvalues(micron)
Washedwool
y=1.2286x- 42.889
R²=0.789010
5
10
15
20
25
30
35
40
45
40 45 50 55 60 65 70
OFD
ADia(micron)
OonSAdirectvalues(micron)
Greasywoll
-
OFDA & OonSA calibration – values comparison
• More variation with greasy then washed • With washed good comparison obtained throughout grades – some highly
deviated• More sets of samples would improve accuracy
Results
0
5
10
15
20
25
30
35
40
45
36 44 46 48 50 54 56 58 60 62 64 70 80
Woolfiberdiameter(m
icron)
Grade
Washedwool
OFDA OonSA
0
5
10
15
20
25
30
35
40
45
36 40 44 46 48 50 54 56 58 60 62 64 70 80
Woolfiberdiameter(m
icron)
Grade
GreasywoolOFDA OonSA
-
• Developed NDSU OonSA plugin produces acceptable results
• Cost of the developed system excluding the computer is well than $200 (scanner + supplies) Field of view better for scanner
• Preparation time for laying out sample is slightly longer than OFDA
• Processing time (scan time + analysis) is 1 minute 15 seconds –high RAM computers will have better speed
• Variation exists even with OFDA compared to grade chart values
• OonSA is rich in features yet simple to operate (textual and graphical outputs + data storage)
• A user manual was developed
Conclusions
-
Status of Budget
• $6,800 returned to Wool Council
– We didn’t think buying computers and
scanners was necessary.
– USB key is the only deliverable needed.
-
Future Developments
• Additional field testing
• Calibration routine already included – but need to be
developed for user application
• Users scan known wools and develop their calibration –
hence better measurements (a standard similar to OFDA)
• With user-defined calibration, OonSA can work
irrespective of (i) type of scanner or imaging device –
(camera, microscope), (ii) type of sample – extended to
other type of fibers (e.g., mohair)
-
The Good:
• R2 between 0.79 and
0.90, depending on
washed vs. clean
• Relatively user
friendly
• $200 cost – just a
scanner
• Easily upgraded to
address
improvements
• 1.5 minutes to run a
sample
The Limitations:
• Version 1 is a proof of
concept, additional
testing is needed.
• Sample layout is still
very important for
accuracy.
• It measures a small
area, not a whole
staple.
-
Where could it be used best?
• Vocational Education instructors
• Youth activities and livestock shows
• Small ram sales that don’t have access to the
OFDA units
• Measuring the affects of nutrition/stress
throughout the staple length (research?)
• Extension Agents
• In my opinion, its best use if it is kept cheap
(free?) and highly available.
-
What’s next?
• Wool Council needs to decide if they like
the product and want to continue
development – we are interested in
upgrading it.
• We will publish the results. It will be open
access at that point – no patents.
• Marketing? You tell us how you would like
to proceed. We are not marketers.
-
Questions