R - counting the number of objects in an image using bwlabel - r

Question

i'm rather new to image analysis in R and was wondering how i can assess the number of individual plants within a picture such as this one:
I thought of converting the picture to a black/white picture and then using the bwlabel function to count the number of objects within the picture like this:
R<-R(image)
G<-G(image)
B<-B(image)
ExGreen<-2*G-R-B
plot(ExGreen)
ExGreen<-threshold(ExGreen,thr = "auto",approx=FALSE,adjust=1)
plot(ExGreen)
ExGreen<-clean(ExGreen,10)
plot(ExGreen)
labels=bwlabel(ExGreen)
max(labels)
However, I'm running into the issue that my white colored potato plants do not always form 1 contiguous unity.
I was therefore wondering whether there is some option to connect the white pixels which are very close to each other or whether it is possible to draw a circle around every potato plant and then using the bwlabel function...
Or is there any other option to solve my problem.
Thanks in advance!

Answer 1

I was not aware that R has a package imager for image processing that already has a good deal of builtin functions for solving this problem. Thanks for pointing me to it through this interesting question.
Here is my solution (beware that some thresholds are hard coded and thus not scale invariant!):
library(imager)
image <- load.image("plants.png")
R<-R(image); G<-G(image); B<-B(image)
ExGreen<-2*G-R-B
plot(ExGreen)
# blur before thresholding to fill some gaps
ExGreen <- isoblur(ExGreen, 3)
ExGreen <- threshold(ExGreen, thr="auto", approx=FALSE, adjust=1)
plot(ExGreen)
# split into connected component and keep only large CCs
ccs <- split_connected(ExGreen)
largeccs <- purrr::keep(ccs, function(x) {sum(x) > 800})
plot(add(largeccs))
# count CCs
cat(sprintf("Number of large CCs: %i\n", length(largeccs)))