Top Banner
E. Belin 1 , D. Rousseau 2 , D. Demilly 3 and C. Dürr 4 1 : Laboratoire d'Ingénierie des Systèmes Automatisés (LISA), Université d'Angers, Angers, France. 2 : Université de Lyon, CREATIS, Lyon, France. 3 : GEVES - Station Nationale d'Essais de Semences (SNES), Beaucouzé, France. 4 : INRA - Institut de Recherche en Horticulture et Semences (IRHS), Beaucouzé, France. Non invasive computer vision tools to monitor seedling elongation INTERNATIONAL SEED TESTING ASSOCIATION (ISTA) ANNUAL MEETING 2012 VENLO –THE NETHERLANDS
12

Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

Sep 17, 2018

Download

Documents

vohanh
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

E. Belin1, D. Rousseau2, D. Demilly3 and C. Dürr4

1 : Laboratoire d'Ingénierie des Systèmes Automatisés (LISA), Université d'Angers, Angers, France.2 : Université de Lyon, CREATIS, Lyon, France.

3 : GEVES - Station Nationale d'Essais de Semences (SNES), Beaucouzé, France. 4 : INRA - Institut de Recherche en Horticulture et Semences (IRHS), Beaucouzé, France.

Non invasive computer vision tools to monitor seedling elongation

INTERNATIONAL SEED TESTING ASSOCIATION (ISTA) ANNUAL MEETING 2012VENLO – THE NETHERLANDS

Page 2: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

A multi-disciplinary collaboration

Development of an automated phenotyping of seeds and seedlings within the Federative Research Institute in Angers (www.ifrquasav-angers.org)

Collaboration with research teams in:

• plant sciences

• information sciences

within the Phenotic research project:

www.istia.univ-angers.fr/LISA/PHENOTIC

2/16

Page 3: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

Context: elongation phase

Monitoring the elongation of 2parts of seedlings during growth

-hypocotyl-radicle

Currently:-visible light = actinic-experts notations

⇒ Outperform currentmeasurements withnon invasive computervision tools

Hypocotyl

Radicle

Cotyledons

Page 4: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

Visible images acquisition device

backlighting system

Petri dish with seedlings

cameraresolution: 3200x2400 pixels

t = 40ht = 0h t = 80hEx.: monitoring of Medicago truncatula seedlings during 80h

Page 5: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

Visible images processing

- based on gray-levels images- segmentation by active contour detection

radiclehypocotyl

radicle

Automated extraction of organs curve length

- based on biological prior knowledge. Separation ofhypocotyl and radicle = line done by the germination point

- curve length computed on skeletonof organs

Page 6: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

Results for visible images processing- Tested on 50 seedlings monitored during 50 hours after apparition of embryonicaxis at freq. 2h = 1250 processed images- Comparison between results computed by algorithm (solid line) and resultsobtained by 3 experts notations (dotted line with errorbars)

Ex.: monitoring of hypocotyl (A,C) and radicle (B,D) for 2 seedlings of Medicago truncatula

Page 7: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

Upgrading previous device with thermal imaging

Multimodal imaging acquisition device

t = 70ht = 40ht = 30h

Ex.: monitoring of Medicago truncatula seedlings during 70h

blotter with seedlings

LED lighting system

infrared cooled cameraspectral range: 2,5 – 5 µmresolution: 320x240

visible camera (for visual diagnostic by experts on images)

E. Belin, D. Rousseau, J. Rojas-Varela, D. Demilly, M.H. Wagner, M.H. Cathala, C. Dürr. “Thermography as non invasivefunctional imaging for monitoring seedling growth”; Computers and Electronics in Agriculture, Vol. 79, n° 2, p. 236-240 (2011).

Page 8: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

Thermal images processing n°1

Ex. : monitoring of 4 seedlingsof Mtr during 70h.

Algorithm developed: demonstrate the feasibility of transposing from visible to dark conditions the monitoring of total curve length of roots elongation seedling

18 seedlings monitored during 70h atfreq. 2h30 + apparition of embryonic axis35h after sowing ≈ 250 images

Page 9: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

Observations of multimodal images

Specific contrast on thermal imaging

No influence of the blotter.No contact between root (radicle + hypocotyl) withblotter: specific contrast on thermal images

⇒ Separation: radicle / hypocotyl ?

Page 10: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

- Algorithm developed to extractradicle and hypocotyl curve lengthfrom the separation of specificcontrast.

- Registration of thermal and visibleimages.

- Comparison with notations of 3experts on 250 images.

Thermal images processing n°2 (1/2)

Page 11: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

• Results in good agreement / according to the relative deviation(errorbars) of experts notations• Thermal imaging may be predictive: hypocotyl seen before expertsnotations (panel A)

• Physiological hypothesis: fluxes exchanges during heterotrophgrowth

Thermal images processing n°2 (2/2)

Ex.: monitoring of hypocotyl and radicle of 1 seedling of Mtr during 70h.

Page 12: Non invasive computer vision tools to monitor seedling ... · good agreement between manual measurements and fully automated measurements. ... MH. Wagner, S. Besson, S. Ducournau,

Conclusion & perspectives

2 non invasive computers vision tools to perform organs segmentationof seedlings during elongation.

Focus on thermal imaging device:good agreement between manual measurements and fully automatedmeasurements.

Main interest of thermal imaging: working in dark conditions.

Questions about the functional origin of thermal contrast in seedlingsduring elongation.

These tools will be used in different research projects:AKER, Convigour, plant KBBE,…