development of an leaf tracking algorithm on corn plants

36
Development of an leaf tracking algorithm on corn plants Speaker: Stefan Schwartz

Upload: lemnatec

Post on 15-Mar-2016

216 views

Category:

Documents


3 download

DESCRIPTION

Plant Phenomics leaf tracking

TRANSCRIPT

Page 1: Development of an leaf tracking algorithm on corn plants

Formatvorlage des Untertitelmasters durch Klicken bearbeiten

life science research

high-throughput screening

plant functional genomics

ecotoxicology

high-content screening

cell analysis

colony counting

drug discovery

quality control

Development of an leaf tracking algorithm on corn plants

Speaker: Stefan Schwartz

Page 2: Development of an leaf tracking algorithm on corn plants

challanges and aim

idea and algorithm

implementation

results

discussion

future prospects

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 3: Development of an leaf tracking algorithm on corn plants

challanges and aim

idea and algorithm

implementation

results

discussion

future prospects

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 4: Development of an leaf tracking algorithm on corn plants

challanges and aim

idea and algorithm

implementation

results

discussion

future prospects

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 5: Development of an leaf tracking algorithm on corn plants

challanges and aim

idea and algorithm

implementation

results

discussion

future prospects

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 6: Development of an leaf tracking algorithm on corn plants

challanges and aim

idea and algorithm

implementation

results

discussion

future prospects

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 7: Development of an leaf tracking algorithm on corn plants

challanges and aim

idea and algorithm

implementation

results

discussion

future prospects

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 8: Development of an leaf tracking algorithm on corn plants

challanges and aim

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 9: Development of an leaf tracking algorithm on corn plants

challanges and aim

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 9

LemnaTec develops hardware- and software solutions for automated high throuput phenotyping

main aim is the digital quantification of plant features in

and beyond human perception

software development includes

hardware control

database management

image processing

Page 10: Development of an leaf tracking algorithm on corn plants

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 10

challanges and aim

Page 11: Development of an leaf tracking algorithm on corn plants

challanges and aim

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 11

many information on the outer shape of a plant are available

would it possible to identify the structure of a plant?

how could leaves be counted?

Page 12: Development of an leaf tracking algorithm on corn plants

challanges and aim

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 12

to keep track on single leaves the inner structure of the plant needs to be identified and the plant has to be segmented

leaf 1

leaf 2

leaf 3

leaf 4

leaf 5 leaf 6

Page 13: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 14: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 14

skeletonization algorithms reduce surfaces to single lines

skeletonization is already used to identify objects (e.g. fingerprinting)

graph theory provides a big amount of algorithms for path analysis

Image Skeleton Graph

Segmented Plant

Page 15: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 15

the idea of the algorithm can therefore be divided into 5 steps:

create graph

simplify graph

interpret graph I (stem)

interpret graph II (leaves)

vectorise and tabulate data

Page 16: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 16

skeletons contain all points that have at least two border points with the same distance to the point itself

the skeleton is divided into joints and bones

the outer shape of the original object is still avaiable

Page 17: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 17

a graph is used to show connections between certain points

it contains edges and vertices

its structure is saved in a matrix

multiple efficient path tracking algorithms are described for graphs

Page 18: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 18

the skeleton will be transformed into a graph

bones become edges

joints become vertices

J2

J1

B1

B2 B3

B4

B5

B6

V1

V2

V5

V7

V4

V6

V3

E1

E2 E3

E4

E5

E6

Page 19: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 19

the sekelton and therefore the created graph can contain artifacts

these artifacts can massivly influence later analysis steps

gap

false edge

Page 20: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 20

the mean edge length is calculated 𝑙𝑒𝑖 ≔ 𝑙𝑒𝑛𝑔𝑡𝑕 𝑜𝑓 𝑒𝑑𝑔𝑒 𝑒𝑖

𝑙 = 1

|𝔼| 𝑙𝑒𝑖

|𝔼|

𝑖 = 0

for each vertex with just one outgoing edge:

gaps will be filled if: 𝑣𝑖𝑣𝑗 ≤ 𝜚 ∙ 𝑙

edges will be deleted if: 𝑙𝑒𝑘 ≤ 𝑙 ∙ 𝜑

Page 21: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 21

the stem is a straight path in the center of the plant

the initial and end direction of each edge is calculated

𝛾𝑎 = ∢ 𝑒 , 𝑎 𝛾𝑏 = ∢ 𝑒 , 𝑏 𝑚𝑖𝑡 0 ≤ 𝛾𝑎 , 𝛾𝑏 ≤ 𝜋

Page 22: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 22

the Dijkstra-Algorithm calculates a path between an initial and an end vertex with the lowest costs

the initial vertex kann be manually configured

the costs of an edge can be determined with the defined directions

stem ends in the highest branch

𝑪𝒐𝒔𝒕𝑺 𝒆𝒊 = 𝜸𝒂 −𝝅

𝟐+ 𝜸𝒃 −

𝝅

𝟐

Page 23: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 23

leaves begin on a leaf tip and grow on an even path to the stem

the costs of each edge were calculated by using the angle of the intersection to the next leaf

𝐶𝑜𝑠𝑡𝐿 𝑒𝑖 , 𝑒𝑗 = 𝜋 − (𝛾𝑎,𝑒𝑗 + 𝛾𝑏,𝑒𝑖)

Page 24: Development of an leaf tracking algorithm on corn plants

idea and algorithm

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 24

complex structure of leaves is represented by using three characteristic points

each leaf is numbered by the Point S

α = ∢ 𝑒 , 𝑆𝑇 β = ∢ 𝑆𝑇 , 𝐸𝑇 γ = ∢ 𝑒 , 𝐸𝑇 𝑚𝑖𝑡 0 ≤ α, β, γ ≤ 𝜋

Page 25: Development of an leaf tracking algorithm on corn plants

implementation

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 26: Development of an leaf tracking algorithm on corn plants

implementation

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 26

integrated in existing software (LemnaGrid)

graphical User Interface for development of image processing chains

Page 27: Development of an leaf tracking algorithm on corn plants

results

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 28: Development of an leaf tracking algorithm on corn plants

results

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 28

Page 29: Development of an leaf tracking algorithm on corn plants

results

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 29

Page 30: Development of an leaf tracking algorithm on corn plants

results

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 30

Page 31: Development of an leaf tracking algorithm on corn plants

results

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 31

Page 32: Development of an leaf tracking algorithm on corn plants

discussion

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 33: Development of an leaf tracking algorithm on corn plants

discussion

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 33

leaves and stem have been identified

analysis without mistakes can not be guaranted

quality criteria improves results

algorithm was tested on a big corn database

using the quality criteria decreased the number of results by 7%

Page 34: Development of an leaf tracking algorithm on corn plants

future prospects

Overview

Date: 2/15/2011 Speaker: Stefan Schwartz

Page 35: Development of an leaf tracking algorithm on corn plants

Future prospects

Date: 2/15/2011 Speaker: Stefan Schwartz Slide 35

algorithm could compare two images

creating 3D models to avoid length changes due to rotation effects

devolving the algorithm to other plants

Page 36: Development of an leaf tracking algorithm on corn plants

Visit our website www.lemnatec.com

Thank you for your attention!