Package to calculate Canopy Openness Index (COI) from canopy images
require("devtools")
devtools::install_github("Edbbioeco/coiR")coiR package is downloaded now, so we can library it. Additionally, we
library the following packages:
-
terra: import images files as rasters;
-
purrr: create loops for multiple operations;
-
ggplot2: create elegant graphs for imported images;
-
tidyterra: visualize RGB raster.
Important
Before load packages, make sure they are previously downloaded.
library(coiR)Carregando pacotes exigidos: terra
terra 1.8.93
Carregando pacotes exigidos: tidyterra
Anexando pacote: 'tidyterra'
O seguinte objeto é mascarado por 'package:stats':
filter
Carregando pacotes exigidos: dplyr
Anexando pacote: 'dplyr'
Os seguintes objetos são mascarados por 'package:terra':
intersect, union
Os seguintes objetos são mascarados por 'package:stats':
filter, lag
Os seguintes objetos são mascarados por 'package:base':
intersect, setdiff, setequal, union
Carregando pacotes exigidos: tidyr
Anexando pacote: 'tidyr'
O seguinte objeto é mascarado por 'package:terra':
extract
Carregando pacotes exigidos: magrittr
Anexando pacote: 'magrittr'
O seguinte objeto é mascarado por 'package:tidyr':
extract
Os seguintes objetos são mascarados por 'package:terra':
extract, inset
Carregando pacotes exigidos: forcats
Carregando pacotes exigidos: ggplot2
Carregando pacotes exigidos: purrr
Anexando pacote: 'purrr'
O seguinte objeto é mascarado por 'package:magrittr':
set_names
Carregando pacotes exigidos: readr
Carregando pacotes exigidos: stringr
Carregando pacotes exigidos: tibble
library(terra)
library(purrr)
library(ggplot2)
library(tidyterra)Now, we need to import our data. Images may be shotten photos, as .png,
.jpg or .jpeg files. first, we informe images directory (files), and
import them using terra::rast() function for every image, throught a
loop with purrr::map() function. Our images (images) are setted as a
list class object.
files <- paste0("cropped-images/imagem", 1:4, ".png")
files[1] "cropped-images/imagem1.png" "cropped-images/imagem2.png"
[3] "cropped-images/imagem3.png" "cropped-images/imagem4.png"
images <- purrr::map(files,
terra::rast)Warning: [rast] unknown extent
Warning: [rast] unknown extent
Warning: [rast] unknown extent
Warning: [rast] unknown extent
names(images) <- paste0("cropped-images/imagem", 1:4, ".png")
images$`cropped-images/imagem1.png`
class : SpatRaster
size : 2971, 2971, 4 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 2971, 0, 2971 (xmin, xmax, ymin, ymax)
coord. ref. :
source : imagem1.png
names : imagem1_1, imagem1_2, imagem1_3, imagem1_4
$`cropped-images/imagem2.png`
class : SpatRaster
size : 2999, 2999, 4 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 2999, 0, 2999 (xmin, xmax, ymin, ymax)
coord. ref. :
source : imagem2.png
names : imagem2_1, imagem2_2, imagem2_3, imagem2_4
$`cropped-images/imagem3.png`
class : SpatRaster
size : 2999, 2999, 4 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 2999, 0, 2999 (xmin, xmax, ymin, ymax)
coord. ref. :
source : imagem3.png
names : imagem3_1, imagem3_2, imagem3_3, imagem3_4
$`cropped-images/imagem4.png`
class : SpatRaster
size : 3000, 3000, 4 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 3000, 0, 3000 (xmin, xmax, ymin, ymax)
coord. ref. :
source : imagem4.png
names : imagem4_1, imagem4_2, imagem4_3, imagem4_4
Next, lets visualize every image, using a purrr::map() loop, through
ggplot and tidyterra::geom_spatraster_rgb() function.
purrr::map(images, function(data){ggplot() + tidyterra::geom_spatraster_rgb(data = data) + theme_void()})<SpatRaster> resampled to 501264 cells.
<SpatRaster> resampled to 501264 cells.
<SpatRaster> resampled to 501264 cells.
<SpatRaster> resampled to 501264 cells.
$`cropped-images/imagem1.png`
$`cropped-images/imagem2.png`
$`cropped-images/imagem3.png`
$`cropped-images/imagem4.png`
For our exemples, lets work in two ways: run for a single and run for multiple images. Lets set first image as a single image.
single_image <- images[[1]]Usely, canopy images are shotten photos, square images. For our
analysis, we need to crop images into circles. We use
coiR::coir_crop() |function.
single_image |>
coiR::coir_crop()<SpatRaster> resampled to 501264 cells.
class : SpatRaster
size : 2971, 2971, 4 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 2971, 0, 2971 (xmin, xmax, ymin, ymax)
coord. ref. :
source(s) : memory
varname : imagem1
names : imagem1_1, imagem1_2, imagem1_3, imagem1_4
min values : 0, 0, 0, 0
max values : 246, 255, 255, 255
And we also can analyse multiple images from an one shot, using
purrr::map() loop.
purrr::map(images, coiR::coir_crop)<SpatRaster> resampled to 501264 cells.
<SpatRaster> resampled to 501264 cells.
<SpatRaster> resampled to 501264 cells.
<SpatRaster> resampled to 501264 cells.
$`cropped-images/imagem1.png`
class : SpatRaster
size : 2971, 2971, 4 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 2971, 0, 2971 (xmin, xmax, ymin, ymax)
coord. ref. :
source(s) : memory
varname : imagem1
names : imagem1_1, imagem1_2, imagem1_3, imagem1_4
min values : 0, 0, 0, 0
max values : 246, 255, 255, 255
$`cropped-images/imagem2.png`
class : SpatRaster
size : 2999, 2999, 4 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 2999, 0, 2999 (xmin, xmax, ymin, ymax)
coord. ref. :
source(s) : memory
varname : imagem2
names : imagem2_1, imagem2_2, imagem2_3, imagem2_4
min values : 0, 0, 0, 0
max values : 255, 255, 255, 255
$`cropped-images/imagem3.png`
class : SpatRaster
size : 2999, 2999, 4 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 2999, 0, 2999 (xmin, xmax, ymin, ymax)
coord. ref. :
source(s) : memory
varname : imagem3
names : imagem3_1, imagem3_2, imagem3_3, imagem3_4
min values : 0, 0, 0, 0
max values : 255, 255, 255, 255
$`cropped-images/imagem4.png`
class : SpatRaster
size : 3000, 3000, 4 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 3000, 0, 3000 (xmin, xmax, ymin, ymax)
coord. ref. :
source(s) : memory
varname : imagem4
names : imagem4_1, imagem4_2, imagem4_3, imagem4_4
min values : 0, 0, 0, 5
max values : 255, 255, 255, 255
Our next step is to binarize our images. We use coiR::coir_binarize()
function. To avoid replot canopy image, we use plot = FALSE argument
at coiR::coir_crop() function. To facilite our analysis, lets use pipe
(|>) to conect functions output.
single_image |>
coiR::coir_crop(plot = FALSE) |>
coiR::coir_binarize()<SpatRaster> resampled to 501264 cells.
class : SpatRaster
size : 2971, 2971, 1 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 2971, 0, 2971 (xmin, xmax, ymin, ymax)
coord. ref. :
source(s) : memory
name : variavel
min value : 0
max value : 1
As previously made, we can binarize multiple images, making a function
in our purrr::map() loop.
purrr::map(images, function(images){coiR::coir_crop(data = images,
plot = FALSE) |>
coiR::coir_binarize()})<SpatRaster> resampled to 501264 cells.
<SpatRaster> resampled to 501264 cells.
<SpatRaster> resampled to 501264 cells.
<SpatRaster> resampled to 501264 cells.
$`cropped-images/imagem1.png`
class : SpatRaster
size : 2971, 2971, 1 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 2971, 0, 2971 (xmin, xmax, ymin, ymax)
coord. ref. :
source(s) : memory
name : variavel
min value : 0
max value : 1
$`cropped-images/imagem2.png`
class : SpatRaster
size : 2999, 2999, 1 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 2999, 0, 2999 (xmin, xmax, ymin, ymax)
coord. ref. :
source(s) : memory
name : variavel
min value : 0
max value : 1
$`cropped-images/imagem3.png`
class : SpatRaster
size : 2999, 2999, 1 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 2999, 0, 2999 (xmin, xmax, ymin, ymax)
coord. ref. :
source(s) : memory
name : variavel
min value : 0
max value : 1
$`cropped-images/imagem4.png`
class : SpatRaster
size : 3000, 3000, 1 (nrow, ncol, nlyr)
resolution : 1, 1 (x, y)
extent : 0, 3000, 0, 3000 (xmin, xmax, ymin, ymax)
coord. ref. :
source(s) : memory
name : variavel
min value : 0
max value : 1
Finally, we get COI index. We use coiR::coir_index() for get that
value. To avoid replot cropped and binarized images, we set
plot = FALSE for both coiR::coir_crop() and coiR::coir_binarize()
functions.
single_image |>
coiR::coir_crop(plot = FALSE) |>
coiR::coir_binarize(plot = FALSE) |>
coiR::coir_index()[1] 0.51
As previously made, we can also do for multiple images, making a
function in purrr::map() function.
purrr::map(images, function(images){coiR::coir_crop(data = images,
plot = FALSE) |>
coiR::coir_binarize(plot = FALSE) |>
coiR::coir_index()})$`cropped-images/imagem1.png`
[1] 0.51
$`cropped-images/imagem2.png`
[1] 0.54
$`cropped-images/imagem3.png`
[1] 0.31
$`cropped-images/imagem4.png`
[1] 0.31
Additionally, we can set all COI values into a vector, to make a
dataframe. First, we make a null vector (vector_index <- c()). Next,
we build a function, hyper declaring a new vector_index with <<-
instead <-. For more values details, we set round = 3 argument in
coiR::coir_index() functions, to set decimal places. All in a
purrr::map() loop.
Note
A good practice to assist loops for images in coiR package is store them into a contained folder, where images file names got a pattern.
vector_index <- c()
multiple_index <- function(images, verbose = FALSE){
index <- images |>
coiR::coir_crop(plot = FALSE) |>
coiR::coir_binarize(plot = FALSE) |>
coiR::coir_index(round = 3)
vector_index <<- c(vector_index, index)
}
purrr::map(images, multiple_index)$`cropped-images/imagem1.png`
[1] 0.507
$`cropped-images/imagem2.png`
[1] 0.507 0.540
$`cropped-images/imagem3.png`
[1] 0.507 0.540 0.310
$`cropped-images/imagem4.png`
[1] 0.507 0.540 0.310 0.308
vector_index[1] 0.507 0.540 0.310 0.308
Finally, we make a data frame with those values.
Note
That data frame can be exported for different file formats, such as .xlsx or .csv. That practice assists save analysis results.
df_index <- tibble::tibble(id = images |> names(),
Index = vector_index)
df_index# A tibble: 4 × 2
id Index
<chr> <dbl>
1 cropped-images/imagem1.png 0.507
2 cropped-images/imagem2.png 0.54
3 cropped-images/imagem3.png 0.31
4 cropped-images/imagem4.png 0.308