Creation of big.matrix object from RasterBrick

Question

In order to perform a kmean cluster analysis on large raster datasets I am attempting to convert my RasterBrick object into a big.matrix object using the brick function, but when I read the .grd file back into R all the information is lost.

library(raster)
library(bigmemory)
library(biganalytics)

#initialize raster
one <- raster(matrix(rnorm(400), 20, 20))
two <- raster(matrix(rnorm(400), 20, 20))
three <- raster(matrix(rnorm(400), 20, 20))

#save brick object as .grd file
brick(one, two, three, filename = "test")

#read .grd file in as big.matrix
big_matrix <- as.big.matrix("test.grd", type = "double")

#check dimensions
dim(big_matrix)

#perform kmeans
bigkmeans(big_matrix, 3)

I can see the the .grd and .gri files in my directory, but I can't figure out how to read them back in, or how to feed the .grd file into bigkmean function. Any idea how I might do this?

Answer 1

Example data

library(raster)
library(bigmemory)

b <- brick(system.file("external/rlogo.grd", package="raster"))

If the file is not that large, you could do

x <- as.big.matrix(values(b))

Otherwise, here is a function you can use.

r2bm <- function(from, filename="") {
    b <- big.matrix(ncell(from), nlayers(from), backingfile=filename )
    nc <- ncol(from)
    tr <- blockSize(from)
    for (i in 1:tr$n) {
        start <- ((tr$row[i]-1) * nc) + 1
        end <- start + (tr$nrows[i] * nc) - 1
        b[start:end, ] <- getValues(from, row=tr$row[i], nrows=tr$nrows[i])
    }
    b
}

Now use it

y <- r2bm(b, "bg.dat")