使用sf包按ID分组从点创建多线。

Question

我有一个点的集合，其中包含一个LINEID和一个ID_SEQ。LINEID确定唯一的线路ID，而ID_SEQ确定线路ID中点的顺序。

我想将点转换为线路，按ID_SEQ排序并按LINEID分组。

使用sp包，我能够实现所需的结果，但我想使用sf包来完成。我在这里缺少什么？

以下是一些虚拟数据，通过sp函数说明了所需结果，并尝试使用sf做相同的事情，但绘图显示出不同的结果。

library(sp)
library(sf)

sfpoints <- {
structure(list(LINEID = c(4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 4L, 10L, 10L, 
10L), ID_SEQ = c(10L, 11L, 12L, 13L, 16L, 30L, 31L, 32L, 33L, 
34L, 35L, 36L, 37L, 38L, 39L, 40L, 41L, 42L, 43L, 44L, 45L, 46L, 
1L, 2L, 3L), geometry = structure(list(structure(c(15.423568, 
47.06248), class = c("XY", "POINT", "sfg")), structure(c(15.423644, 
47.062523), class = c("XY", "POINT", "sfg")), structure(c(15.423691, 
47.062553), class = c("XY", "POINT", "sfg")), structure(c(15.423712, 
47.06257), class = c("XY", "POINT", "sfg")), structure(c(15.423716, 
47.062576), class = c("XY", "POINT", "sfg")), structure(c(15.423712, 
47.062588), class = c("XY", "POINT", "sfg")), structure(c(15.423731, 
47.062595), class = c("XY", "POINT", "sfg")), structure(c(15.423779, 
47.062626), class = c("XY", "POINT", "sfg")), structure(c(15.423835, 
47.062664), class = c("XY", "POINT", "sfg")), structure(c(15.423879, 
47.062714), class = c("XY", "POINT", "sfg")), structure(c(15.423897, 
47.062767), class = c("XY", "POINT", "sfg")), structure(c(15.423862, 
47.062828), class = c("XY", "POINT", "sfg")), structure(c(15.423783, 
47.062897), class = c("XY", "POINT", "sfg")), structure(c(15.423681, 
47.062973), class = c("XY", "POINT", "sfg")), structure(c(15.423564, 
47.06306), class = c("XY", "POINT", "sfg")), structure(c(15.423437, 
47.063164), class = c("XY", "POINT", "sfg")), structure(c(15.42331, 
47.06327), class = c("XY", "POINT", "sfg")), structure(c(15.423186, 
47.063385), class = c("XY", "POINT", "sfg")), structure(c(15.423062, 
47.063496), class = c("XY", "POINT", "sfg")), structure(c(15.422941, 
47.063602), class = c("XY", "POINT", "sfg")), structure(c(15.422821, 
47.063717), class = c("XY", "POINT", "sfg")), structure(c(15.422699, 
47.063824), class = c("XY", "POINT", "sfg")), structure(c(15.422518, 
47.061687), class = c("XY", "POINT", "sfg")), structure(c(15.422617, 
47.06179), class = c("XY", "POINT", "sfg")), structure(c(15.422717, 
47.061893), class = c("XY", "POINT", "sfg"))), class = c("sfc_POINT", 
"sfc"), precision = 0, bbox = structure(c(15.422518, 47.061687, 
15.423897, 47.063824), .Names = c("xmin", "ymin", "xmax", "ymax"
), class = "bbox"), crs = structure(list(epsg = NA_integer_, 
proj4string = NA_character_), .Names = c("epsg", "proj4string"
), class = "crs"), n_empty = 0L)), .Names = c("LINEID", "ID_SEQ", 
"geometry"), row.names = c(NA, -25L), class = c("sf", "data.frame"
), sf_column = "geometry", agr = structure(c(NA_integer_, NA_integer_
), .Names = c("LINEID", "ID_SEQ"), .Label = c("constant", "aggregate", 
"identity"), class = "factor"))}

par(mfrow=c(1,3))

## SP - way
tstssp <- as(sfpoints, "Spatial")
tstssp <- SpatialLines(lapply(split(tstssp, tstssp$LINEID), function(x) 
  Lines(list(Line(coordinates(x))), x$LINEID[1L])))
plot(tstssp, col=1:2, lwd=3, main="SP-Desired Result")


## SF - way ???
tst <- sfpoints %>% 
  group_by(LINEID) %>% 
  st_coordinates() %>%
  st_linestring()
plot(st_geometry(tst), col=1:2 , main="SF-Wrong Result")

tst <- sfpoints %>% 
  group_by(LINEID) %>%
  summarise() %>%
  st_cast("LINESTRING")
plot(st_geometry(tst), col=1:2, main="SF-Wrong Result")

- SeGa

---

我几天前遇到了非常类似的问题。在 sf 的 Github 上有一些问题报告：https://github.com/r-spatial/sf/issues/325 和 https://github.com/r-spatial/sf/issues/321。`st_union` 似乎不会关注您数据框中的顺序，而是根据哪些点彼此最接近进行排序。 - camille

在我的情况下，我最终通过ID将数据框拆分，并利用了gtfsr包中的shapes_df_as_sfg，然后将其全部绑定在一起并转换为线串。不幸的是，我无法将其适应于您的sf。 - camille

感谢您的回复。我已经看到了那些Github问题，但它们对我的情况没有帮助。我唯一能做的就是改变为sp-structure，将点分割成线，然后再转换回sf。但我没有看到st_union取最近的下一个点，因为这会在图3中产生正确的结果，所以我仍然想知道它如何将点合并在一起，从而输出这种锯齿状结构。 - SeGa

我之前说的是最近点，这是我的错误。我认为实际上是按照经度值递增排序，这就是为什么会出现锯齿形的原因。 - camille

那可以解释这个奇怪的形状。可惜我无法传递ID_SEQ告诉他正确的顺序。 - SeGa

Answer 1

如 Edzer 在这篇讨论中所解释的，您需要在summarise函数中提供参数do_union = FALSE:

tst <- sfpoints %>% 
  group_by(LINEID) %>%
  summarise(do_union = FALSE) %>%
  st_cast("LINESTRING")
plot(st_geometry(tst), col=1:2)

- sebdalgarno

---

太好了！我读了那个问题好几遍，一直缺少关键部分“do_union = FALSE”。非常感谢你！ - SeGa