Measurement of Plant Characteristics Using Digital Images

E
TWC/26/21 Rev.
ORIGINAL: English
DATE: August 25, 2008
INTERNATIONAL UNION FOR THE PROTECTION OF NEW VARIETIES OF PLANTS
GENEVA
TECHNICAL WORKING PARTY ON AUTOMATION AND
COMPUTER PROGRAMS
Twenty-Sixth Session
Jeju, Republic of Korea, September 2 to 5, 2008
MEASUREMENT OF PLANT CHARACTERISTICS USING DIGITAL IMAGES
Document prepared by experts from United Kingdom
TWC/26/21 Rev.
page 2
Measurement of plant
characteristics using digital
images
Adrian Roberts, Graham Horgan, Alec Mann
Niall Green, Tom Christie, George Campbell
TWC/26, Republic of Korea 2008
1
Why automate measurements?
•
Reduce costs
– Select characteristics
– Often benefit when can measure several characteristics on same image
– Need to optimise process
•
•
Improve consistency
Develop new measurements
•
Images can be reviewed long after normal measurement time
– Stored images provide testbed
– Quality assurance
– Evidence in case of challenge
•
Produce images for reference collection management
TWC/26, Republic of Korea 2008
2
TWC/26/21 Rev.
page 3
UK Vegetable DUS Centre
Collaboration between Science and Advice for Scottish
Agriculture (SASA) and Biomathematics & Statistics
Scotland (BioSS) since 2000
– Now in routine use:
•
•
•
•
Pea: pod, leaflet, stipule, petiole, peduncle
Parsnip: root
Broad bean: leaves
After validation against manual measurement and cost-benefit study
– Other crops and characteristics in development
• Brassica cotyledons, pods; Broad bean seeds …
– Software developed - Imagin
• Fortran routines accessed by a Visual Basic Graphical User Interface
• Demo later
TWC/26, Republic of Korea 2008
3
Types of characteristics
• Measurements of overall size
– e.g. length/width, area
• Specific dimensions
– e.g. curvature and see later
• Complex measurements
– e.g. dentation, curvature
• Colour-based measurements
– Needs careful set up
– In Scotland, leaving as visual scoring
• New characteristics
– Need to go through process of approval for use in UPOV guideline
TWC/26, Republic of Korea 2008
4
TWC/26/21 Rev.
page 4
Outline of process
• Take digital image of sample
– Costly bit (collection, preparation and layout of plant parts)
– Require optimisation – compromise between quality and cost
– Could do in the field (in situ) or in controlled conditions
• Images - digitised RGB
– for each pixel, have coords and R, G & B values
• Need to identify and label objects (positioning and colour
can be useful)
• Identify outlines and landmarks (set orientation helps)
• Make measurements
• Rescale measurements (coin)
TWC/26, Republic of Korea 2008
5
Tips
• Consistent photographic conditions
– Within year but also from year to year
– Oblique lighting
• Black background
– Helps in identification of objects and tends to reduce photographic
problems
• Objects should not touch or overlap
– Heavy glass sheet on stipules/leaflets to flatten leaves (small
measurement error associated with leaf folding when flattened)
• Scaling object
• QA issues in labelling and naming
– barcodes
TWC/26, Republic of Korea 2008
6
TWC/26/21 Rev.
page 5
TWC/26/21 Rev.
page 6
Pea automated stipule measurements
A
B
Character
Existing
C-D Stipule: width
Existing
E
B-J
G
UPOV
A-C Stipule: length
Stipule Area
New
Stipule: length from axil to tip
New
M-K Stipule: Length of lobe below axil New
H
F
Petiole: length is measured in the
Existing
same image as the stipule
J
Stipule: margin dentation
M
K
L
C
TWC/26, Republic of Korea 2008
J-L, J-K, K-L, E-G, F-H, G-H
D
10
R&D
R&D
TWC/26/21 Rev.
page 7
Imagin
Imagin GUI (ADM)
Reads images from
directories
Image analysis
Routines (GWH)
Handles options
Process image and
generate results.
Passes images and
options to image
analysis routines
Written in Fortran77
Presents results
Written in Visual
Basic
TWC/26, Republic of Korea 2008
12 34.1 11.23 58.9 ...
11
Average image
Area of current development
• Have many images from same variety
• Want to represent by single shape – concept
of average shape
• Have developed for parsnip roots and
stipules so far
TWC/26, Republic of Korea 2008
12
TWC/26/21 Rev.
page 8
TWC/26, Republic of Korea 2008
13
Demo of Imagin
TWC/26, Republic of Korea 2008
14
[End of document]