Currently I have two data.frames, one of polygons (poly.x, poly.y, enum)
and one of points (pt.x, pt.y)
where enum
is the id of the polygon. I am trying to determine which points belong to which polygons so I get a data.frame of (pt.x, pt.y, enum)
.
My first attempt uses point.in.polygon
from the sp
package and lapply
functions to find which polygon(s) the point belongs to. While my code works, it takes a long time on large data sets.
My second attempt uses over
also from the sp
package, cobbled together from questions on gis stackexchange. While it is much faster, I cannot seem to get the correct output from over
as it is a dataframe of 1
s and NA
s.
Below I've included a simplified working example ( npoly
can be changed to test the speed of different methods) as well as my working attempt using sp::point.in.polygon
and nonsensical output from my sp::over
attempt. I'm not fussed which method I end up using as long as it's fast.
Any help would be much appreciated!
#-------------------------------------------
# Libraries
library(ggplot2) # sample plots
library(dplyr) # bind_rows(), etc
library(sp) # spatial data
# Sample data
npoly = 100
# polygons
localpolydf <- data.frame(
x = rep(c(0, 1, 1, 0), npoly) + rep(0:(npoly-1), each = 4),
y = rep(c(0, 0, 1, 1), npoly),
enum = rep(1:npoly, each = 4))
# points
offsetdf <- data.frame(
x = seq(min(localpolydf$x) - 0.5, max(localpolydf$x) + 0.5, by = 0.5),
y = runif(npoly*2 + 3, 0, 1))
# Sample plot
ggplot() +
geom_polygon(aes(x, y, group = enum),
localpolydf, fill = NA, colour = "black") +
geom_point(aes(x, y), offsetdf)
#-------------------------------------------
# Dplyr and lapply solution for point.in.polygon
ptm <- proc.time() # Start timer
# create lists
offsetlist <- split(offsetdf, rownames(offsetdf))
polygonlist <- split(localpolydf, localpolydf$enum)
# lapply over each pt in offsetlist
pts <- lapply(offsetlist, function(pt) {
# lapply over each polygon in polygonlist
ptpoly <- lapply(polygonlist, function(poly) {
data.frame(
enum = poly$enum[1],
ptin = point.in.polygon(pt[1,1], pt[1,2], poly$x, poly$y))
})
ptpoly <- bind_rows(ptpoly) %>% filter(ptin != 0)
if (nrow(ptpoly) == 0) return(data.frame(x = pt$x, y = pt$y, enum = NA, ptin = NA))
ptpoly$x = pt$x
ptpoly$y = pt$y
return(ptpoly[c("x", "y", "enum", "ptin")])
})
pts_apply <- bind_rows(pts)
proc.time() - ptm # end timer
#-------------------------------------------
# Attempted sp solution for over
ptm <- proc.time() # Start timer
# Split the dataframe into a list based on enum and then remove enum from df in the list
polygonlist <- split(localpolydf, localpolydf$enum)
polygonlist <- lapply(polygonlist, function(x) x[,c("x", "y")])
# Convert the list to Polygon, then create a Polygons object
polygonsp <- sapply(polygonlist, Polygon)
polygonsp <- Polygons(polygonsp, ID = 1)
polygonsp <- SpatialPolygons(list(polygonsp))
plot(polygonsp)
# Convert points to coordinates
offsetps <- offsetdf
coordinates(offsetps) <- ~x+y
points(offsetps$x, offsetps$y)
# Determine polygons points are in
pts_sp <- over(offsetps, polygonsp)
proc.time() - ptm # end timer
#===========================================
# Output
# Apply: point.in.polygon
> head(pts_apply)
x y enum ptin
1 -0.5 0.2218138 NA NA
2 4.0 0.9785541 4 2
3 4.0 0.9785541 5 2
4 49.0 0.3971479 49 2
5 49.0 0.3971479 50 2
6 49.5 0.1177206 50 1
user system elapsed
4.434 0.002 4.435
# SP: over
> head(pts_sp)
1 2 3 4 5 6
NA 1 1 NA 1 NA
user system elapsed
0.048 0.000 0.047
After having another look, I realised Roman did pts_sp == 1
because I only had 1 ID for all of my squares, ie when I did ID = 1
.
Once I fixed that, I was able to a column with ID = enum
. To handle points in multiple polygons I can use returnList = TRUE
and add additional lines to convert the list to a data.frame but it isn't necessar here.
# Attempted sp solution
ptm <- proc.time() # Start timer
# Split the dataframe into a list based on enum and then remove enum from df in the list
polygonlist <- split(localpolydf, localpolydf$enum)
# Convert the list to Polygon, then create a Polygons object
polygonsp <- sapply(polygonlist, function(poly){
Polygons(list(Polygon(poly[, c("x", "y")])), ID = poly[1, "enum"])
})
# polygonsp <- Polygons(polygonsp, ID = 1)
polygonsp <- SpatialPolygons(polygonsp)
plot(polygonsp)
# Convert points to coordinates
offsetps <- offsetdf
coordinates(offsetps) <- ~x+y
points(offsetps$x, offsetps$y)
# Determine polygons points are in
pts_sp <- over(offsetps, polygonsp)
pts_sp <- data.frame(
x = offsetps$x, y = offsetps$y,
enum = unique(localpolydf$enum)[pts_sp])
proc.time() - ptm # end timer
An alternative to using over
is to use sf::intersection
as the sf package is becoming more and more popular.
Getting the data into sf objects took me a little bit of work but if you are working with external data you can just read in with st_read
and it will already be in the correct form.
Here is how to approach:
library(tidyverse)
library(sf)
# convert into st_polygon friendly format (all polygons must be closed)
# must be a nicer way to do this!
localpoly <- localpolydf %>% split(localpolydf$enum) %>%
lapply(function(x) rbind(x,x[1,])) %>%
lapply(function(x) x[,1:2]) %>%
lapply(function(x) list(as.matrix(x))) %>%
lapply(function(x) st_polygon(x))
# convert points into sf object
points <- st_as_sf(offsetdf,coords=c('x','y'),remove = F)
#convert polygons to sf object and add id column
polys <- localpoly %>% st_sfc() %>% st_sf(geom=.) %>%
mutate(id=factor(1:100))
#find intersection
joined <- polys %>% st_intersection(points)
# Sample plot
ggplot() + geom_sf(data=polys) +
geom_sf(data=joined %>% filter(id %in% c(1:10)),aes(col=id)) +
lims(x=c(0,10))
Note that to use geom_sf at the time of writing you will need to install the development version of ggplot.
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.