Plant IMage Analysis

Tiago Olivoto

2021-05-13

Getting started

pliman (plant image analysis) is designed to analyze plant images, especially related to leaf analysis. You provide color palettes, tell pliman what each one represents, and it takes care of the details. The package will help you to:

Leaf area

The function leaf_area() can be used to computes the leaf area using an image with leaves and a template with a known area. A general linear model (binomial family) fitted to the RGB values is used to first separate the leaves and template from the background and then the leaves from the template. The leaf area is then calculated for each leaf based on the pixel area.

In the following example, the area of five three leaves will be computed using a reference template with a known area of 4 cm\(^2\).

library(pliman)
# |========================================================|
# | Tools for Plant Image Analysis (pliman 0.2.0)          |
# | Author: Tiago Olivoto                                  |
# | Type 'vignette('pliman_start')' for a short tutorial   |
# | Visit 'https://bit.ly/3eL0dF3' for a complete tutorial |
# |========================================================|
img <- image_import(image_pliman("la_pattern.JPG"))
leaf <- image_import(image_pliman("la_leaf.jpg"))
tmpl <- image_import(image_pliman("la_temp.jpg"))
background <- image_import(image_pliman("la_back.jpg"))
image_combine(img, leaf, tmpl, background)



# Computes the leaf area
area <- 
leaf_area(img = img,
          img_leaf = leaf,
          img_template = tmpl,
          img_background = background,
          area_template = 4,
          text_col = "white")

get_measures(area)
# -----------------------------------------
# Total leaf area  : 142.931 
# Average leaf area: 23.822 
# -----------------------------------------
#   id        x        y      area perimeter radius_mean radius_min radius_max
# 1  1 510.8398 194.8310 33.675574       750   130.93825  112.84830  174.98737
# 2  2 206.7294 201.3974 27.063816       707   119.27683   93.26143  170.78366
# 3  3 123.4206 557.3541 30.701037       787   128.71937   95.70023  189.29542
# 4  4 622.9502 585.3997 39.408991       864   143.38098  108.58618  203.37775
# 5  5 364.7605 508.5782  8.081735       376    64.26213   51.24084   86.01652

Counting crop grains

Here, we will count the grains in the image soybean_touch.png. This image has a cyan background and contains 30 soybean grains that touch with each other.

soy <- image_import(image_pliman("soybean_touch.jpg"))
image_show(soy)


# Count the objects in the image
grains <- count_objects(soy)
# 
# --------------------------------------------
# Number of objects: 30 
# --------------------------------------------
#  statistics       area   perimeter
#         min  1366.0000  117.000000
#        mean  2057.3667  146.600000
#         max  2445.0000  158.000000
#          sd   230.5574    8.406073
#         sum 61721.0000 4398.000000

# Draws the object id (by default)
plot_measures(grains)

Disease severity

The function symptomatic_area() is used to compute the percentage of symptomatic leaf area in a sample or entire leaf based on provided color palettes samples. A general linear model (binomial family) fitted to the RGB values is used to segment the lesions from the healthy leaf. If a pallet of background is provided, the function takes care of the details to isolate it before computing the number and area of lesions.

The next example computes the symptomatic area of a soybean leaf. The proportion of the healthy and symptomatic areas is given as the proportion of the total leaf area, after segmenting the leaf from the background (blue).

img <- image_import(image_pliman("sev_leaf.jpg"))
healthy <- image_import(image_pliman("sev_healthy.jpg"))
symptoms <- image_import(image_pliman("sev_sympt.jpg"))
background <- image_import(image_pliman("sev_back.jpg"))
image_combine(img, healthy, symptoms,background)


# Computes the symptomatic area
symptomatic_area(img = img,
                 img_healthy = healthy,
                 img_symptoms = symptoms,
                 img_background = background,
                 show_image = TRUE)

#    healthy symptomatic
# 1 89.11082    10.88918

RGB values for objects

The function objects_rgb() can be used to get the Red Green and Blue (RGB) for objects in an image. Here, the RGB for

img <- image_import(image_pliman("soy_green.jpg"))
# Segment the foreground (grains) using the normalized blue index
# Shows the average value of the blue index in each object
rgb <- objects_rgb(img, marker = "index")


# Draw the object id
image_show(img)
plot_measures(rgb)


# plot the distribution of RGB values of each object
plot(rgb)

Getting help