4

I have a very big dataset of points (SpatialPoints object) over a line (SpatialLines object). I need to sort these points following the line direction.

Note: the number of points and coordinates that build the SpatialLines object are not necessarily the same of the SpatialPoints object.

I tried this (How to make a SpatialLines object from undefined - but known - spatial sequence of points in R?) using only the SpatialPoints object but the dataset is too big and my computer collapse (can't allocate all the data to memory).

An example using a small dataset sample:

# Load packages
library('sp')

# Define projection
epsg.32721 <- "+proj=utm +zone=21 +south +datum=WGS84 +units=m +no_defs"

# Load data points
x <- c(794801.3, 795014.1, 795055.7, 795051.1, 795070.9, 794885.1, 795102.9, 795131.9, 795127.4, 795192.6, 795196.3, 795219.3, 795283.2, 795307.2, 795307.2, 795307.7, 795327.7, 795330.2, 795360.6, 795355.7, 794779.8)
y <- c(6194137, 6194211, 6194192, 6194192, 6194202, 6194210, 6194207, 6194217, 6194217, 6194231, 6194234, 6194275, 6194279, 6194713, 6194713, 6194714, 6194795, 6194335, 6194869, 6194865, 6194109)

# Create SpatialPoints object
spatialPointsObject <- SpatialPoints(coords = cbind(x,y), proj4string = CRS(epsg.32721))

# Load data lines
x2 <- c(794779.8, 794790.6, 794795.7, 794801.3, 794810.2, 794815.9, 794828.5, 794838.2, 794846.4, 794844.0, 794850.5, 794858.6, 794869.3, 794877.9, 794886.3, 794895.5, 794904.8, 794914.9, 794926.2, 794938.3, 794948.1, 794958.7, 794969.6, 794978.4, 794987.1, 794999.1, 795008.2, 795017.6, 795025.5, 795032.2, 795045.8, 795055.7, 795064.3, 795073.5, 795083.1, 795093.2, 795103.4, 795112.6, 795122.2, 795131.9, 795142.3, 795152.4, 795162.3, 795168.3, 795178.9, 795187.8, 795196.3, 795200.5, 795207.6, 795213.0, 795219.3, 795268.8, 795275.3, 795283.2, 795289.9, 795308.4, 795313.5, 795320.6, 795330.2, 795344.0, 795344.7, 795350.1, 795354.6, 795367.8, 795372.8, 795377.6, 795367.6, 795367.5, 795377.5, 795372.5, 795372.5, 795367.6, 795367.6, 795367.6, 795367.5, 795367.5, 795367.5, 795372.6, 795372.7, 795377.6, 795377.6, 795377.6, 795377.5, 795377.5, 795372.5, 795372.6, 795368.5, 795363.0, 795362.5, 795362.1, 795363.1, 795307.2, 795315.1, 795318.8, 795325.1, 795325.9, 795328.8, 795334.4, 795331.9, 795334.2, 795338.3, 795346.1, 795353.1, 795365.9, 795373.4, 795389.3, 795388.9, 795390.4, 795394.7, 795400.2, 795410.4, 795411.0, 795416.5, 795413.8, 795422.3, 795423.4, 795424.4, 795427.2, 795427.0, 795431.5)    
y2 <- c(6194109, 6194113, 6194126, 6194137, 6194141, 6194142, 6194146, 6194155, 6194160, 6194183, 6194191, 6194197, 6194199, 6194204, 6194210, 6194215, 6194221, 6194227, 6194232, 6194227, 6194227, 6194227, 6194227, 6194222, 6194222, 6194215, 6194210, 6194211, 6194207, 6194206, 6194192, 6194192, 6194200, 6194202, 6194207, 6194207, 6194207, 6194212, 6194217, 6194217, 6194222, 6194227, 6194222, 6194222, 6194224, 6194228, 6194234, 6194242, 6194264, 6194269, 6194275, 6194270, 6194278, 6194279, 6194291, 6194294, 6194309, 6194317, 6194335, 6194350, 6194360, 6194369, 6194378, 6194393, 6194410, 6194426, 6194441, 6194452, 6194463, 6194475, 6194487, 6194497, 6194507, 6194517, 6194526, 6194536, 6194546, 6194549, 6194556, 6194567, 6194577, 6194588, 6194605, 6194616, 6194623, 6194637, 6194654, 6194671, 6194672, 6194683, 6194692, 6194713, 6194736, 6194744, 6194751, 6194765, 6194814, 6194819, 6194829, 6194836, 6194844, 6194850, 6194863, 6194874, 6194889, 6194903, 6194910, 6194922, 6194930, 6194940, 6194947, 6194956, 6194968, 6194980, 6194986, 6194997, 6195006, 6195016, 6195036, 6195051)

# Create SpatialLines object
spatialLinesObject <- SpatialLines(LinesList = list(Lines(slinelist = list(Line(coords = cbind(x2,y2))), ID = "1")), proj4string = CRS(epsg.32721))

# Plot SpatialPoints + SpatialLines + coordinates(SpatialLines)
plot(spatialLinesObject, xlab = "Longitude", ylab = "Latitude", main = "SpatialPoints + SpatialLines + coordinates(SpatialLines)")
plot(SpatialPoints(coordinates(spatialLinesObject)), pch = 19, col = "black", add = TRUE, cex = 0.4)
plot(spatialPointsObject, col = "green", add = TRUE)
text(coordinates(spatialPointsObject), labels = seq_along(spatialPointsObject), cex = 0.7, pos = 3)
box()

I need to get a SpatialPoints object where the first point should be the Point 21, second point should be Point 1, third point should be Point 6 and so on: see this

3

I could answer my question traveling through every line segment in the SpatialLines object and finding the intersected points from the SpatialPoints object. I made a function that is all explained by comments.

# Load packages
library('rgeos')
library('data.table')

# Vector norm
norm_vec <- function(x) sqrt(sum(x^2))    

# OrderPointsAlongLine function
OrderPointsAlongLine <- function(spatialLinesObject, spatialPointsObject) {

  lineCoords = as.data.frame(coordinates(spatialLinesObject))
  lineCoordsLength = nrow(lineCoords)

  # Make a buffer 
  spatialPointsObjectBuffer = gBuffer(spgeom = spatialPointsObject, width = 1, byid = TRUE)

  # Aux list and data frame
  cellsAndDistances = list()
  cd = data.frame('cell' = NA, 'dist' = NA)

  # Go sequentially through every line segment in the SpatialLine object looking for which SpatialPoints are intersected
  for(i in 1:(lineCoordsLength - 1)) {

    # Unommented: show iteration number
    # cat(paste("\n iteration", i, sep = ""))

    # Line segment points
    p0 = lineCoords[i,]
    p1 = lineCoords[i+1,]

    # Build line segment
    SL = SpatialLines(LinesList = list(Lines(Line(coords = rbind(p0,p1)), ID = paste(i, sep = ""))), proj4string = CRS(epsg.32721))

    # Find intersected points with line segment
    intersect = which(gIntersects(spgeom1 = spatialPointsObjectBuffer, spgeom2 = SL, byid = TRUE))

    # Points intersected
    p2 = spatialPointsObject[intersect,]

    p2Length = nrow(p2@coords)

    # If there is almost one point intersected
    if(p2Length != 0) {

      # If there are more than one intersected points
      for(j in 1:p2Length) {

        # Commented: show sub-iteration number
        # cat(paste("\n sub-iteration", j, sep = ""))

        cd[j,] = data.frame('cell' = intersect[j], 'dist' = norm_vec(coordinates(p2)[j,] - p0))

      }
    }

    # Order by distance
    data.table::setorder(cd, dist) 
    cellsAndDistances[[i]] = cd

  }

  # Bind list into one data farme
  cellsAndDistancesBind = do.call("rbind", cellsAndDistances)

  # Remove duplicated
  cellsAndDistancesBind <- cellsAndDistancesBind[-which(duplicated(cellsAndDistancesBind$cell)),]

  # Return
  return(cellsAndDistancesBind)

}

# Call function
orderedPoints <- OrderPointsAlongLine(spatialLinesObject = spatialLinesObject, spatialPointsObject = spatialPointsObject)

# Plots
plot(spatialLinesObject, xlab = "Longitude", ylab = "Latitude", main = "", col = "green")
plot(spatialPointsObject[orderedPoints$cell,], add = TRUE, pch = 19, cex = 0.5)
text(coordinates(spatialPointsObject[orderedPoints$cell,]), labels = seq_along(spatialPointsObject[orderedPoints$cell,]), cex = 0.75, pos = 1)

ordered points along line

Improvements are welcome!

  • I think your function call should be OrderPointsAlongLine instead of SortPointsAlongLine. – SeGa Jun 29 '18 at 9:00
2

Here is the same function, but only using Simple Features. Although it is apparently less performant and takes almost twice the time.

library(sf)
OrderPointsAlongLineSF <- function(spatialLinesObject, spatialPointsObject) {

  lineCoords = data.table(st_coordinates(spatialLinesObject))
  lineCoordsLength = nrow(lineCoords)

  # Make a buffer 
  spatialPointsObjectBuffer = st_buffer(spatialPointsObject, dist = 1)

  # Aux list and data frame
  cellsAndDistances = vector("list", lineCoordsLength-1)
  cd = data.frame('cell' = NA, 'dist' = NA)

  # Go sequentially through every line segment in the SpatialLine object looking for which SpatialPoints are intersected
  for(i in 1:(lineCoordsLength - 1)) {
    # Line segment points
    p0 = lineCoords[i, c("X", "Y")]
    p1 = lineCoords[i+1, c("X", "Y")]


    SF <- st_as_sf(rbind(p0,p1), coords = c("X", "Y")) %>% 
      summarise(do_union = FALSE) %>%
      st_cast("LINESTRING") 
    SF$id = paste(i, sep = "")
    st_crs(SF) <- st_crs(spatialLinesObject)

    # Find intersected points with line segment
    intersect = which(st_intersects(spatialPointsObjectBuffer, SF, FALSE))

    # Points intersected
    p2 = spatialPointsObject[intersect,]

    p2Length = nrow(p2)

    # If there is almost one point intersected
    if (p2Length != 0) {
      # If there are more than one intersected points
      for(j in 1:p2Length) {
        cd[j,] = data.frame('cell' = intersect[j], 'dist' = norm_vec(st_coordinates(p2)[j,] - p0))
      }
    }

    # Order by distance
    data.table::setorder(cd, dist) 
    cellsAndDistances[[i]] = cd
  }

  # Bind list into one data farme
  cellsAndDistancesBind = do.call("rbind", cellsAndDistances)

  # Remove duplicated
  cellsAndDistancesBind <- cellsAndDistancesBind[-which(duplicated(cellsAndDistancesBind$cell)),]

  # Return
  return(cellsAndDistancesBind)
}

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.