Whole plant segmentation

This script allows to reproduce Figure 5 that shows plant representations at different development stages (one to eight weeks after sowing) for (a) top and (b) side images, and time courses of the number of pixels corresponding to plants extracted from side and top views (c)

The following materials are available at

A folder named Maize whole plant image dataset contains:

1. Image dataset: image dataset of side and top RGB images of a single plant that can be used in the segmentation pipeline (https://github.com/openalea/eartrack)

2. Segmented image dataset: output images of the segmentation pipeline

3. Image analysis features: csv file containing all image analysis features from the image dataset provided above

Figure 5a,b

Plant representations at different developmental stages (one to eight weeks after sowing) for (a) top and (b) side images

#Required R packages
if (!require("EBImage")) install.packages("EBImage")

par(mfrow=c(2,5))
#top images
topimage= list.files("./FIG5/top/")
for(image in topimage){
  imageiday = readImage(paste("./FIG5/top/", image,sep=""))
  display(imageiday, method='raster')
}
#side images
sideimage= list.files("./FIG5/side/")
for(image in sideimage){
  imageiday = readImage(paste("./FIG5/side/", image,sep=""))
  display(imageiday, method='raster')
}

Get output data from segmentation pipeline

This exemple is based on a csv file that contains image analysis features from one single plant

#Load csv file
data_output <- read.csv("./segmentationdata.csv")
knitr::kable(head(data_output))

datetime	viewtypeid	angle	convex_hull_perimeter	width	convex_hull_area	object_sum_area	height_over_pot	height_under_pot	height	Sowing	Day	DAS
14/04/2017 01:06	1	0	219.1609	71	2231.5	1577	NA	NA	71	96	14/04/2017	8
14/04/2017 01:06	2	0	303.7828	50	3066.0	1713	138	0	137	96	14/04/2017	8
14/04/2017 01:06	2	330	278.4993	31	1549.0	1171	134	0	133	96	14/04/2017	8
14/04/2017 01:06	2	300	273.5317	32	2371.5	1382	131	0	130	96	14/04/2017	8
14/04/2017 01:06	2	270	295.2435	55	3674.5	1515	132	0	131	96	14/04/2017	8
14/04/2017 01:06	2	240	313.4547	70	4645.5	1663	133	0	132	96	14/04/2017	8

Figure 5c

Time courses of the number of pixels corresponding to plants extracted from side and top views.

par(bg=NA)
plot(object_sum_area/10^3~DAS, type='b', data=data_output[data_output$viewtypeid==1,], las=1, col=2, pch=1, ylab='Pixel number (10^3)', xlab="Days after sowing")
points(object_sum_area/10^3~DAS, type='b', data=data_output[data_output$angle==90,], col=2, pch=0, ylab='', las=1)

legend("topleft", c("Top view","Side 90° view"), pch=c(1,0), col=c(2,2), bty='n')