注意：

Question

我想使用R叠加一些空间点和多边形，以便为这些点分配我所考虑的地理区域的某些属性。

我通常要做的是使用over软件包中的命令sp。我的问题是，我正在处理全球范围内发生的大量地理参考事件，在某些情况下（尤其是在沿海地区），经度和纬度的组合略微超出了国家/地区边界。这里是一个基于此very good question的可复制示例。

## example data
set.seed(1)
library(raster)
library(rgdal)
library(sp)
p <- shapefile(system.file("external/lux.shp", package="raster"))
p2 <- as(0.30*extent(p), "SpatialPolygons")
proj4string(p2) <- proj4string(p)
pts1 <- spsample(p2-p, n=3, type="random")
pts2<- spsample(p, n=10, type="random")
pts<-rbind(pts1, pts2)

## Plot to visualize
plot(p, col=colorRampPalette(blues9)(12))
plot(pts, pch=16, cex=.5,col="red", add=TRUE)

# overlay
pts_index<-over(pts, p)

# result
pts_index

#>     ID_1       NAME_1 ID_2           NAME_2 AREA
#>1    NA         <NA> <NA>             <NA>   NA
#>2    NA         <NA> <NA>             <NA>   NA
#>3    NA         <NA> <NA>             <NA>   NA
#>4     1     Diekirch    1         Clervaux  312
#>5     1     Diekirch    5            Wiltz  263
#>6     2 Grevenmacher   12     Grevenmacher  210
#>7     2 Grevenmacher    6       Echternach  188
#>8     3   Luxembourg    9 Esch-sur-Alzette  251
#>9     1     Diekirch    3          Redange  259
#>10    2 Grevenmacher    7           Remich  129
#>11    1     Diekirch    1         Clervaux  312
#>12    1     Diekirch    5            Wiltz  263
#>13    2 Grevenmacher    7           Remich  129

有没有一种方法可以赋予over函数某种公差，以便同时捕获非常靠近边界的点？

注意：

在this之后，我可以将丢失的点分配给最近的多边形，但这并不完全是我想要的。

编辑：最近的邻居解决方案

#adding lon and lat to the table
pts_index$lon<-pts@coords[,1]
pts_index$lat<-pts@coords[,2]

#add an ID to split and then re-compose the table 
pts_index$split_id<-seq(1,nrow(pts_index),1)
#filtering out the missed points

library(dplyr)
library(geosphere)
missed_pts<-filter(pts_index, is.na(NAME_1))
pts_missed<-SpatialPoints(missed_pts[,c(6,7)],proj4string=CRS(proj4string(p)))

#find the nearest neighbors' characteristics
n <- length(pts_missed)
nearestID1 <- character(n)
nearestNAME1 <- character(n)
nearestID2 <- character(n)
nearestNAME2 <- character(n)
nearestAREA <- character(n)

for (i in seq_along(nearestID1)) {
  nearestID1[i] <- as.character(p$ID_1[which.min(dist2Line (pts_missed[i,], p))])
  nearestNAME1[i] <- as.character(p$NAME_1[which.min(dist2Line (pts_missed[i,], p))])
  nearestID2[i] <- as.character(p$ID_2[which.min(dist2Line (pts_missed[i,], p))])
  nearestNAME2[i] <- as.character(p$NAME_2[which.min(dist2Line (pts_missed[i,], p))])
  nearestAREA[i] <- as.character(p$AREA[which.min(dist2Line (pts_missed[i,], p))])
}
missed_pts$ID_1<-nearestID1
missed_pts$NAME_1<-nearestNAME1
missed_pts$ID_2<-nearestID2
missed_pts$NAME_2<-nearestNAME2
missed_pts$AREA<-nearestAREA

#missed_pts have now the characteristics of the nearest poliygon
#bringing now everything toogether
pts_index[match(missed_pts$split_id, pts_index$split_id),] <- missed_pts
pts_index<-pts_index[,-c(6:8)]

pts_index

       ID_1       NAME_1 ID_2           NAME_2 AREA
1     1     Diekirch    4          Vianden   76
2     1     Diekirch    4          Vianden   76
3     1     Diekirch    4          Vianden   76
4     1     Diekirch    1         Clervaux  312
5     1     Diekirch    5            Wiltz  263
6     2 Grevenmacher   12     Grevenmacher  210
7     2 Grevenmacher    6       Echternach  188
8     3   Luxembourg    9 Esch-sur-Alzette  251
9     1     Diekirch    3          Redange  259
10    2 Grevenmacher    7           Remich  129
11    1     Diekirch    1         Clervaux  312
12    1     Diekirch    5            Wiltz  263
13    2 Grevenmacher    7           Remich  129

这与@Gilles在他的答案中提出的输出完全相同。我只是想知道是否有比这更有效的东西。

Answer 1

这是我尝试使用 sf 的尝试。如果您想盲目地将多边形要素连接到与其最近邻点的点，只需用st_join调用join = st_nearest_feature

library(sf)

# convert data to sf
pts_sf = st_as_sf(pts)
p_sf = st_as_sf(p)

# this is enough for joining polygon attributes to points from their nearest neighbor
st_join(pts_sf, p_sf, join = st_nearest_feature)

如果您希望能够设置一些公差，以使与该公差相距较远的点不会加入任何多边形属性，则我们需要创建自己的联接函数。

st_nearest_feature2 = function(x, y, tolerance = 100) {
  isec = st_intersects(x, y)
  no_isec = which(lengths(isec) == 0)

  for (i in no_isec) {
    nrst = st_nearest_points(st_geometry(x)[i], y)
    nrst_len = st_length(nrst)
    nrst_mn = which.min(nrst_len)
    isec[i] = ifelse(as.vector(nrst_len[nrst_mn]) > tolerance, integer(0), nrst_mn)
  }

  unlist(isec)

}

st_join(pts_sf, p_sf, join = st_nearest_feature2, tolerance = 1000)

这可以按预期工作，即，当您将tolerance设置为零时，您将获得与over相同的结果，对于较大的值，您将从上方接近st_nearest_feature的结果。

Answer 2

示例数据-

set.seed(1)
library(raster)
library(rgdal)
library(sp)
p <- shapefile(system.file("external/lux.shp", package="raster"))
p2 <- as(0.30*extent(p), "SpatialPolygons")
proj4string(p2) <- proj4string(p)
pts1 <- spsample(p2-p, n=3, type="random")
pts2<- spsample(p, n=10, type="random")
pts<-rbind(pts1, pts2)

## Plot to visualize
plot(p, col=colorRampPalette(blues9)(12))
plot(pts, pch=16, cex=.5,col="red", add=TRUE)

使用sf和nngeo软件包的解决方案-

library(nngeo)

# Convert to 'sf'
pts = st_as_sf(pts)
p = st_as_sf(p)

# Spatial join
p1 = st_join(pts, p, join = st_nn)
p1

## Simple feature collection with 13 features and 5 fields
## geometry type:  POINT
## dimension:      XY
## bbox:           xmin: 5.795068 ymin: 49.54622 xmax: 6.518138 ymax: 50.1426
## epsg (SRID):    4326
## proj4string:    +proj=longlat +datum=WGS84 +no_defs
## First 10 features:
##   ID_1       NAME_1 ID_2           NAME_2 AREA                  geometry
## 1     1     Diekirch    4          Vianden   76 POINT (6.235953 49.91801)
## 2     1     Diekirch    4          Vianden   76 POINT (6.251893 49.92177)
## 3     1     Diekirch    4          Vianden   76  POINT (6.236712 49.9023)
## 4     1     Diekirch    1         Clervaux  312  POINT (6.090294 50.1426)
## 5     1     Diekirch    5            Wiltz  263  POINT (5.948738 49.8796)
## 6     2 Grevenmacher   12     Grevenmacher  210 POINT (6.302851 49.66278)
## 7     2 Grevenmacher    6       Echternach  188 POINT (6.518138 49.76773)
## 8     3   Luxembourg    9 Esch-sur-Alzette  251 POINT (6.116905 49.56184)
## 9     1     Diekirch    3          Redange  259 POINT (5.932418 49.78505)
## 10    2 Grevenmacher    7           Remich  129 POINT (6.285379 49.54622)

显示连接哪些多边形和点的图-

# Visuzlize join
l = st_connect(pts, p, dist = 1)
plot(st_geometry(p))
plot(st_geometry(pts), add = TRUE)
plot(st_geometry(l), col = "red", lwd = 2, add = TRUE)

编辑：

# Spatial join with 100 meters threshold
p2 = st_join(pts, p, join = st_nn, maxdist = 100)
p2
## Simple feature collection with 13 features and 5 fields
## geometry type:  POINT
## dimension:      XY
## bbox:           xmin: 5.795068 ymin: 49.54622 xmax: 6.518138 ymax: 50.1426
## epsg (SRID):    4326
## proj4string:    +proj=longlat +datum=WGS84 +no_defs
## First 10 features:
##   ID_1       NAME_1 ID_2           NAME_2 AREA                  geometry
## 1    NA         <NA> <NA>             <NA>   NA POINT (6.235953 49.91801)
## 2    NA         <NA> <NA>             <NA>   NA POINT (6.251893 49.92177)
## 3     1     Diekirch    4          Vianden   76  POINT (6.236712 49.9023)
## 4     1     Diekirch    1         Clervaux  312  POINT (6.090294 50.1426)
## 5     1     Diekirch    5            Wiltz  263  POINT (5.948738 49.8796)
## 6     2 Grevenmacher   12     Grevenmacher  210 POINT (6.302851 49.66278)
## 7     2 Grevenmacher    6       Echternach  188 POINT (6.518138 49.76773)
## 8     3   Luxembourg    9 Esch-sur-Alzette  251 POINT (6.116905 49.56184)
## 9     1     Diekirch    3          Redange  259 POINT (5.932418 49.78505)
## 10    2 Grevenmacher    7           Remich  129 POINT (6.285379 49.54622)

Answer 3

我认为您不能在over或其他常见的相交算法中添加“公差”。通过缓冲多边形，可以增加一些公差，但是某些点可能会落入两个不同的多边形中。

一种可能性是在区域多边形之外的点周围创建缓冲区，将这些缓冲区与多边形相交，计算面积，并为每个点仅保留最大面积的线。相对于您建议的方法（查找最接近的多边形），这种方法的优势在于您不必计算所有多边形的距离。

可能还有更直接的可能性...

这里是一个使用sf来操纵空间对象的示例，但是您当然可以使用sp和rgeos来做同样的事情。
一种困难是找到正确级别的“容差”（缓冲区的大小）。在这里，我使用2公里的公差。

## Your example
set.seed(1)
library(raster)
#> Loading required package: sp
library(rgdal)
library(sp)


p <- shapefile(system.file("external/lux.shp", package="raster"))
p2 <- as(0.30*extent(p), "SpatialPolygons")
proj4string(p2) <- proj4string(p)
pts1 <- spsample(p2-p, n=3, type="random")
pts2<- spsample(p, n=10, type="random")
pts<-rbind(pts1, pts2)

请注意，我在over上的输出与您不同：

over(pts, p)
#>    ID_1       NAME_1 ID_2           NAME_2 AREA
#> 1    NA         <NA> <NA>             <NA>   NA
#> 2    NA         <NA> <NA>             <NA>   NA
#> 3    NA         <NA> <NA>             <NA>   NA
#> 4     1     Diekirch    1         Clervaux  312
#> 5     1     Diekirch    5            Wiltz  263
#> 6     2 Grevenmacher   12     Grevenmacher  210
#> 7     2 Grevenmacher    6       Echternach  188
#> 8     3   Luxembourg    9 Esch-sur-Alzette  251
#> 9     1     Diekirch    3          Redange  259
#> 10    2 Grevenmacher    7           Remich  129
#> 11    1     Diekirch    1         Clervaux  312
#> 12    1     Diekirch    5            Wiltz  263
#> 13    2 Grevenmacher    7           Remich  129

在多边形之外的点上使用缓冲区：

# additional packages needed
library(sf)
library(dplyr)

# transform the sp objects into sf objects and add an ID to the points
pts <- st_as_sf(pts)
pts$IDpts <- 1:nrow(pts)
p <- st_as_sf(p)

# project the data in planar coordinates (here a projection for Luxemburg)
# better for area calculations but maybe not crucial here
pts <- st_transform(pts, crs = 2169)
p <- st_transform(p, crs = 2169)


# intersect the points with the polygons (equivalent to you "over")
pts_index <- st_set_geometry(st_intersection(pts, p), NULL)
#> Warning: attribute variables are assumed to be spatially constant
#> throughout all geometries


# points that are outside the polygons
pts_out <- pts[lengths(st_within(pts, p)) == 0,]
# buffer around these points with a given size
bf <- st_buffer(pts_out, dist = 2000) # distance in meters, here 2km

# intersect these buffers with the polygons and compute their area
bf <- st_intersection(bf, p)
#> Warning: attribute variables are assumed to be spatially constant
#> throughout all geometries
bf$area <- st_area(bf)

# for each point (IDpts), select the line with the highest area
# then drop the geometry columns and transform the result n a data.frame
pts_out <- bf %>% group_by(IDpts) %>% slice(which.max(area)) %>% 
     select(1:6) %>% st_set_geometry(NULL) %>% as.data.frame()

输出：

# Colate the results from the point within polygons and outside polygons
pts_index <- rbind(pts_index, pts_out)
pts_index <- pts_index[order(pts_index$IDpts),]
pts_index
#>    IDpts ID_1       NAME_1 ID_2           NAME_2 AREA
#> 1      1    1     Diekirch    4          Vianden   76
#> 2      2    1     Diekirch    4          Vianden   76
#> 3      3    1     Diekirch    4          Vianden   76
#> 4      4    1     Diekirch    1         Clervaux  312
#> 5      5    1     Diekirch    5            Wiltz  263
#> 6      6    2 Grevenmacher   12     Grevenmacher  210
#> 7      7    2 Grevenmacher    6       Echternach  188
#> 8      8    3   Luxembourg    9 Esch-sur-Alzette  251
#> 9      9    1     Diekirch    3          Redange  259
#> 10    10    2 Grevenmacher    7           Remich  129
#> 11    11    1     Diekirch    1         Clervaux  312
#> 12    12    1     Diekirch    5            Wiltz  263
#> 13    13    2 Grevenmacher    7           Remich  129

R以一定程度的公差覆盖点和多边形

注意：

编辑：最近的邻居解决方案

3 个答案: