Plotting raster images using custom colours in R

Question

This might sound like a strange process, but its the best I can think of to control rasterised colour gradients with respect to discrete objects (points, lines, polygons). I'm 95% there but can't quite plot correctly.

This should illustrate proof of concept:

require(raster)
r = matrix(56:255, ncol=20)         # reds
b = t(matrix(56:255, ncol=10))      # blues
col = matrix(rgb(r, 0, b, max=255), ncol=20) # matrix of colour strings
ras = raster(r)                     # data raster object
extent(ras) = extent(1,200,1,100)   # set extent for aspect
plot(ras, col = col, axes=F, asp=T) # overwrite data with custom colours

Here I want to clip a raster to a triangle and create colour gradient of pixels inside based on their distances to one of the sides. Sorry for length but its the most minimal example I can design.

require(raster); require(reshape2); require(rgeos)
# equilateral triangle
t_s = 100                            # half side
t_h = floor(tan(pi*60/180) * t_s)    # height
corners = cbind(c(0, -t_s, t_s, 0), c(t_h, 0, 0, t_h))
trig = SpatialPolygons(list(Polygons(list(Polygon(corners)),"triangle")))

# line to measure pixel distances to
redline = SpatialLines(list(Lines(Line(corners[1:2,]), ID='redline')))
plot(trig); plot(redline, add=T, col='red', lwd=3)

# create a blank raster and clip to triangle
r = raster(mat.or.vec(nc = t_s*2 + 1, nr = t_h))
extent(r) = extent(-t_s, t_s, 0, t_h)
r = mask(r, trig)
image(r, asp=T)

# extract cell coordinates into d.f.
cells = as.data.frame(coordinates(rasterToPoints(r, spatial=T)))

# calculate distance of each pixel to redline with apply
dist_to_line = function(xy, line){
  point = readWKT(paste('POINT(', xy[1], xy[2], ')'))
  gDistance(point, line) / t_h
}

cells$dists = apply(cells, 1, dist_to_line, line=redline)
cells$cols = rgb(1 - cells$dists, 0, 0)
length(unique(cells$cols)) # count unique colours

# use custom colours to colour triangle pixels
image(r, col = cells$cols, asp=T)
plot(r, col = cells$cols, asp=T)

As you can see the plotting fails to overwrite as in the first example, but the data seems fine. Trying to convert to matrix also fails:

# try convertying colours to matrix
col_ras = acast(cells, y~x, value.var='cols')
col_ras = apply(col_ras, 1, rev) # rotate acw to match r
plot(r, col = col_ras, asp=T)

Very grateful for any assistance on what's going wrong.

---

Edit:

To show Spacedman's plotRGB method:

b = brick(draster, 1-draster, 1-draster)
plotRGB(b, scale=1)
plot(trig, col=NA, border='white', lwd=5, add=T)

Answer 1

Easy way is to go from your points to a spatial pixels data frame to a raster, then do the colour mapping...

Start with:

> head(cells)
           x     y        dists
1  0.0000000 172.5 0.0014463709
2  0.0000000 171.5 0.0043391128
3 -0.9950249 170.5 0.0022523089
4  0.0000000 170.5 0.0072318546
5  0.9950249 170.5 0.0122114004

convert:

> coordinates(cells)=~x+y
> draster = raster(as(cells,"SpatialPixelsDataFrame"))

colourise:

> cols=draster
> cols[!is.na(draster)]= rgb(1-draster[!is.na(draster)],0,0)
> plot(cols, col=cols)

I'm not sure this is the right way to do things though, you might be better off creating an RGB raster stack and using plotRGB if you want fine colour control.