Cargando…

Estimation of the Botanical Composition of Clover-Grass Leys from RGB Images Using Data Simulation and Fully Convolutional Neural Networks

Optimal fertilization of clover-grass fields relies on knowledge of the clover and grass fractions. This study shows how knowledge can be obtained by analyzing images collected in fields automatically. A fully convolutional neural network was trained to create a pixel-wise classification of clover,...

Descripción completa

Detalles Bibliográficos
Autores principales: Skovsen, Søren, Dyrmann, Mads, Mortensen, Anders Krogh, Steen, Kim Arild, Green, Ole, Eriksen, Jørgen, Gislum, René, Jørgensen, Rasmus Nyholm, Karstoft, Henrik
Formato: Online Artículo Texto
Lenguaje:English
Publicado: MDPI 2017
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5751073/
https://www.ncbi.nlm.nih.gov/pubmed/29258215
http://dx.doi.org/10.3390/s17122930
_version_ 1783289866639900672
author Skovsen, Søren
Dyrmann, Mads
Mortensen, Anders Krogh
Steen, Kim Arild
Green, Ole
Eriksen, Jørgen
Gislum, René
Jørgensen, Rasmus Nyholm
Karstoft, Henrik
author_facet Skovsen, Søren
Dyrmann, Mads
Mortensen, Anders Krogh
Steen, Kim Arild
Green, Ole
Eriksen, Jørgen
Gislum, René
Jørgensen, Rasmus Nyholm
Karstoft, Henrik
author_sort Skovsen, Søren
collection PubMed
description Optimal fertilization of clover-grass fields relies on knowledge of the clover and grass fractions. This study shows how knowledge can be obtained by analyzing images collected in fields automatically. A fully convolutional neural network was trained to create a pixel-wise classification of clover, grass, and weeds in red, green, and blue (RGB) images of clover-grass mixtures. The estimated clover fractions of the dry matter from the images were found to be highly correlated with the real clover fractions of the dry matter, making this a cheap and non-destructive way of monitoring clover-grass fields. The network was trained solely on simulated top-down images of clover-grass fields. This enables the network to distinguish clover, grass, and weed pixels in real images. The use of simulated images for training reduces the manual labor to a few hours, as compared to more than 3000 h when all the real images are annotated for training. The network was tested on images with varied clover/grass ratios and achieved an overall pixel classification accuracy of 83.4%, while estimating the dry matter clover fraction with a standard deviation of 7.8%.
format Online
Article
Text
id pubmed-5751073
institution National Center for Biotechnology Information
language English
publishDate 2017
publisher MDPI
record_format MEDLINE/PubMed
spelling pubmed-57510732018-01-10 Estimation of the Botanical Composition of Clover-Grass Leys from RGB Images Using Data Simulation and Fully Convolutional Neural Networks Skovsen, Søren Dyrmann, Mads Mortensen, Anders Krogh Steen, Kim Arild Green, Ole Eriksen, Jørgen Gislum, René Jørgensen, Rasmus Nyholm Karstoft, Henrik Sensors (Basel) Article Optimal fertilization of clover-grass fields relies on knowledge of the clover and grass fractions. This study shows how knowledge can be obtained by analyzing images collected in fields automatically. A fully convolutional neural network was trained to create a pixel-wise classification of clover, grass, and weeds in red, green, and blue (RGB) images of clover-grass mixtures. The estimated clover fractions of the dry matter from the images were found to be highly correlated with the real clover fractions of the dry matter, making this a cheap and non-destructive way of monitoring clover-grass fields. The network was trained solely on simulated top-down images of clover-grass fields. This enables the network to distinguish clover, grass, and weed pixels in real images. The use of simulated images for training reduces the manual labor to a few hours, as compared to more than 3000 h when all the real images are annotated for training. The network was tested on images with varied clover/grass ratios and achieved an overall pixel classification accuracy of 83.4%, while estimating the dry matter clover fraction with a standard deviation of 7.8%. MDPI 2017-12-17 /pmc/articles/PMC5751073/ /pubmed/29258215 http://dx.doi.org/10.3390/s17122930 Text en © 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
spellingShingle Article
Skovsen, Søren
Dyrmann, Mads
Mortensen, Anders Krogh
Steen, Kim Arild
Green, Ole
Eriksen, Jørgen
Gislum, René
Jørgensen, Rasmus Nyholm
Karstoft, Henrik
Estimation of the Botanical Composition of Clover-Grass Leys from RGB Images Using Data Simulation and Fully Convolutional Neural Networks
title Estimation of the Botanical Composition of Clover-Grass Leys from RGB Images Using Data Simulation and Fully Convolutional Neural Networks
title_full Estimation of the Botanical Composition of Clover-Grass Leys from RGB Images Using Data Simulation and Fully Convolutional Neural Networks
title_fullStr Estimation of the Botanical Composition of Clover-Grass Leys from RGB Images Using Data Simulation and Fully Convolutional Neural Networks
title_full_unstemmed Estimation of the Botanical Composition of Clover-Grass Leys from RGB Images Using Data Simulation and Fully Convolutional Neural Networks
title_short Estimation of the Botanical Composition of Clover-Grass Leys from RGB Images Using Data Simulation and Fully Convolutional Neural Networks
title_sort estimation of the botanical composition of clover-grass leys from rgb images using data simulation and fully convolutional neural networks
topic Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5751073/
https://www.ncbi.nlm.nih.gov/pubmed/29258215
http://dx.doi.org/10.3390/s17122930
work_keys_str_mv AT skovsensøren estimationofthebotanicalcompositionofclovergrassleysfromrgbimagesusingdatasimulationandfullyconvolutionalneuralnetworks
AT dyrmannmads estimationofthebotanicalcompositionofclovergrassleysfromrgbimagesusingdatasimulationandfullyconvolutionalneuralnetworks
AT mortensenanderskrogh estimationofthebotanicalcompositionofclovergrassleysfromrgbimagesusingdatasimulationandfullyconvolutionalneuralnetworks
AT steenkimarild estimationofthebotanicalcompositionofclovergrassleysfromrgbimagesusingdatasimulationandfullyconvolutionalneuralnetworks
AT greenole estimationofthebotanicalcompositionofclovergrassleysfromrgbimagesusingdatasimulationandfullyconvolutionalneuralnetworks
AT eriksenjørgen estimationofthebotanicalcompositionofclovergrassleysfromrgbimagesusingdatasimulationandfullyconvolutionalneuralnetworks
AT gislumrene estimationofthebotanicalcompositionofclovergrassleysfromrgbimagesusingdatasimulationandfullyconvolutionalneuralnetworks
AT jørgensenrasmusnyholm estimationofthebotanicalcompositionofclovergrassleysfromrgbimagesusingdatasimulationandfullyconvolutionalneuralnetworks
AT karstofthenrik estimationofthebotanicalcompositionofclovergrassleysfromrgbimagesusingdatasimulationandfullyconvolutionalneuralnetworks