sobel: Apply Sobel filter

In mosscoder/paint2train: What the Package Does (One Line, Title Case)

Description

sobel takes a multi-band raster stack, conducts principal components analysis, then applies a Sobel filter to n retained PCA axes and appends these n bands to the tile.

Usage

sobel(tile, axes = 3, fill_na = TRUE)

Arguments

tile	File path to a mulit-band raster stack.
axes	How many PCA axes to retain and filter.
fill_na	Whether to impute NA cells with mean of neighboring cells.

Value

None. Appends n bands to tile and overwrites appended stack to tile file path.

Examples

library(paint2train)

image_dir <- tempfile()
image_url <- 'https://storage.googleapis.com/mpgranch_data/sample_4band.tif'
download.file(url = image_url, destfile = image_dir)
tdir <- tempdir()
setwd(tdir) 
preproc_dir <- 'preproc_tiles'
dir.create(preproc_dir)

#some test coordinates
xcoords <- c(727495,
             727919)

ycoords <- c(5175339,
             5175408)

coord_mat <- cbind(xcoords, ycoords)

ls <- 30 #how big should the tiles be, this is the side length (in units of data, meters here)
buff <- 5  #buffer in native units of CRS
cores <- ifelse(.Platform$OS.type == 'unix', #how many cores to use for preprocessing
                   parallel::detectCores() - 1,
                   1) 
                   
tile_at_coords(coords = coord_mat,
 len_side = ls,
 buffer = buff,
 out_dir = preproc_dir,
 img = image_dir,
 ncores = cores)
 
t <- list.files(preproc_dir, full.names = TRUE)[1]
ndvi_msavi(tile = t, r_band = 1, nir_band = 4)
sobel(t, axes = 3, fill_na = TRUE)
t_sobels <- stack(t)[[7:9]]
names(t_sobels) <- c('PC1','PC2','PC3')
plot(t_sobels, col = gray.colors(100))

mosscoder/paint2train documentation built on Jan. 21, 2022, 11 a.m.