Convert point shapefile into a polygon which may have holes in the middle using R

Question

I have multiple point shapefiles with 20,000 - 100,000 points and I would like to create an outline/polygon from them using R. There may be holes in the middle of the polygon which I want outlined as well. I have posted pictures of how I want this to look. I have searched a lot for an answer to this using R and haven't found one that has worked. Many other examples I see don't follow the outline as closely as I would like. The polygon that I did get to work came from ArcGIS using the aggregate points (Cartography) tool with a 10 meter distance, but I want to use R.

    ye <- read.csv(ye1,header = F, stringsAsFactors = F)
    ye$gid <- 1
    new_names <- c("easting", "northing","lat","long","yield","moisture","swath","distance","flow","interval","agl","transect","gpstime","utmzone","rmcode","gid")
    names(ye) <- new_names
    
     ye_shp <- SpatialPointsDataFrame(ye[,1:2],data = ye,proj4string = CRS("+proj=utm +zone=14 +datum=NAD83 +units=m +no_defs +ellps=GRS80"))
        

This is an agricultural field with 70,000+ points.

This is what I want the polygon to look like.

Answer 1

Let's see... First create a patchy set of regions a bit like yours by taking a random half of the nc dataset:

set.seed(123)
library(sf)
nc = st_read(system.file("shape/nc.shp", package="sf"))
reg = nc[sample(nrow(nc),nrow(nc)/2),]
reg = st_transform(reg,"EPSG:3857")
plot(reg)

and generate 24,000 points over it:

pts = st_sample(reg, 24000)
plot(pts, add=TRUE)

Now I'll do a merged approximate 5km buffer over the union of the points:

 system.time({b = st_buffer(st_union(pts), dist=5000)})
#   user  system elapsed 
# 34.249   0.404  34.658

which takes 34 seconds on this 4 year old desktop, on a single core. The resulting polygon structure is like this:

 plot(b)

which looks like the thing you are after.

Answer 2

Thanks @Spacedman for the help.

I tried using gBuffer with on the SpatialPointsDataFrame. I initially had a user error by selecting the wrong csv columns for the coordinates. I think gBuffer had issues with coordinate system not using planar coordinates.

Comparing the run times, I went with @Spacedman answer. However, the gBuffer solution by also worked but was slower.

From @Spacedman

    ye <- read.csv(ye1,header = F, stringsAsFactors = F)
    ye_shp <- SpatialPointsDataFrame(ye[,1:2],data = ye,proj4string = CRS("+proj=utm +zone=14 +datum=NAD83 +units=m +no_defs +ellps=GRS80"))
  ye_shp <- st_as_sf(ye_shp)
   system.time({b = st_buffer(st_union(ye_shp), dist=7)})

for a 24k point file
   user   system elapsed 
  28.36    0.31   28.90 

for a 70k point file
  user  system elapsed 
 276.08   14.56  292.56 

gBuffer solution

    ye <- read.csv(ye1,header = F, stringsAsFactors = F)
    ye_shp <- SpatialPointsDataFrame(ye[,1:2],data = ye,proj4string = CRS("+proj=utm +zone=14 +datum=NAD83 +units=m +no_defs +ellps=GRS80"))

  system.time({x <- gBuffer(ye_shp,width = 7)})

for a 24k point file
   user  system  elapsed 
  25.68    0.63   26.53 

for a 70k point file
   user  system elapsed 
 473.14   16.02  493.63