A two-step registration-classification approach to automated segmentation of multimodal images for high-throughput greenhouse plant phenotyping.

Fluorescence imaging Greenhouse plant phenotyping Machine learning Multimodal image alignment Supervised image segmentation Visible light imaging

Journal

Plant methods
ISSN: 1746-4811
Titre abrégé: Plant Methods
Pays: England
ID NLM: 101245798

Informations de publication

Date de publication:
2020
Historique:
received: 19 03 2020
accepted: 30 06 2020
entrez: 17 7 2020
pubmed: 17 7 2020
medline: 17 7 2020
Statut: epublish

Résumé

Automated segmentation of large amount of image data is one of the major bottlenecks in high-throughput plant phenotyping. Dynamic optical appearance of developing plants, inhomogeneous scene illumination, shadows and reflections in plant and background regions complicate automated segmentation of unimodal plant images. To overcome the problem of ambiguous color information in unimodal data, images of different modalities can be combined to a virtual multispectral cube. However, due to motion artefacts caused by the relocation of plants between photochambers the alignment of multimodal images is often compromised by blurring artifacts. Here, we present an approach to automated segmentation of greenhouse plant images which is based on co-registration of fluorescence (FLU) and of visible light (VIS) camera images followed by subsequent separation of plant and marginal background regions using different species- and camera view-tailored classification models. Our experimental results including a direct comparison with manually segmented ground truth data show that images of different plant types acquired at different developmental stages from different camera views can be automatically segmented with the average accuracy of Automated segmentation of arbitrary greenhouse images exhibiting highly variable optical plant and background appearance represents a challenging task to data classification techniques that rely on detection of invariances. To overcome the limitation of unimodal image analysis, a two-step registration-classification approach to combined analysis of fluorescent and visible light images was developed. Our experimental results show that this algorithmic approach enables accurate segmentation of different FLU/VIS plant images suitable for application in fully automated high-throughput manner.

Sections du résumé

BACKGROUND BACKGROUND
Automated segmentation of large amount of image data is one of the major bottlenecks in high-throughput plant phenotyping. Dynamic optical appearance of developing plants, inhomogeneous scene illumination, shadows and reflections in plant and background regions complicate automated segmentation of unimodal plant images. To overcome the problem of ambiguous color information in unimodal data, images of different modalities can be combined to a virtual multispectral cube. However, due to motion artefacts caused by the relocation of plants between photochambers the alignment of multimodal images is often compromised by blurring artifacts.
RESULTS RESULTS
Here, we present an approach to automated segmentation of greenhouse plant images which is based on co-registration of fluorescence (FLU) and of visible light (VIS) camera images followed by subsequent separation of plant and marginal background regions using different species- and camera view-tailored classification models. Our experimental results including a direct comparison with manually segmented ground truth data show that images of different plant types acquired at different developmental stages from different camera views can be automatically segmented with the average accuracy of
CONCLUSION CONCLUSIONS
Automated segmentation of arbitrary greenhouse images exhibiting highly variable optical plant and background appearance represents a challenging task to data classification techniques that rely on detection of invariances. To overcome the limitation of unimodal image analysis, a two-step registration-classification approach to combined analysis of fluorescent and visible light images was developed. Our experimental results show that this algorithmic approach enables accurate segmentation of different FLU/VIS plant images suitable for application in fully automated high-throughput manner.

Identifiants

pubmed: 32670387
doi: 10.1186/s13007-020-00637-x
pii: 637
pmc: PMC7346525
doi:

Types de publication

Journal Article

Langues

eng

Pagination

95

Informations de copyright

© The Author(s) 2020.

Déclaration de conflit d'intérêts

Competing interestsThe authors declare that they have no competing interests.

Références

Trends Plant Sci. 2016 Dec;21(12):989-991
pubmed: 27810146
Sci Rep. 2018 Mar 13;8(1):4465
pubmed: 29535402
Funct Plant Biol. 2013 Oct;40(10):1065-1075
pubmed: 32481174
PeerJ. 2018 Jun 28;6:e5036
pubmed: 29967727
Front Plant Sci. 2015 Jan 20;5:770
pubmed: 25653655
Plant Methods. 2018 Feb 9;14:12
pubmed: 29449872
Trends Plant Sci. 2018 Oct;23(10):883-898
pubmed: 30104148
Plant Methods. 2019 Apr 29;15:44
pubmed: 31168314
Plant Physiol. 2014 Apr 23;165(2):506-518
pubmed: 24760818

Auteurs

Michael Henke (M)

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), OT Gatersleben, Corrensstrasse 3, 06466 Seeland, Germany.

Astrid Junker (A)

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), OT Gatersleben, Corrensstrasse 3, 06466 Seeland, Germany.

Kerstin Neumann (K)

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), OT Gatersleben, Corrensstrasse 3, 06466 Seeland, Germany.

Thomas Altmann (T)

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), OT Gatersleben, Corrensstrasse 3, 06466 Seeland, Germany.

Evgeny Gladilin (E)

Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), OT Gatersleben, Corrensstrasse 3, 06466 Seeland, Germany.

Classifications MeSH