fmesher (version 0.5.0)

fm_nonconvex_hull: Compute an extension of a spatial object

Description

Constructs a potentially nonconvex extension of a spatial object by performing dilation by convex + concave followed by erosion by concave. This is equivalent to dilation by convex followed by closing (dilation + erosion) by concave.

Usage

fm_nonconvex_hull(x, ..., format = "sf", method = "fm")
fm_extensions(
  x,
  convex = -0.15,
  concave = convex,
  ...,
  format = "sf",
  method = "fm"
)
fm_nonconvex_hull_fm(
  x,
  convex = -0.15,
  concave = convex,
  resolution = 40,
  eps = NULL,
  eps_rel = NULL,
  crs = fm_crs(x),
  ...
)
fm_nonconvex_hull_sf(
  x,
  convex = -0.15,
  concave = convex,
  preserveTopology = TRUE,
  dTolerance = NULL,
  crs = fm_crs(x),
  ...
)
# S3 method for sfc
fm_nonconvex_hull(x, ..., format = "sf", method = "fm")
# S3 method for matrix
fm_nonconvex_hull(x, ..., format = "sf", method = "fm")
# S3 method for sf
fm_nonconvex_hull(x, ..., format = "sf", method = "fm")
# S3 method for Spatial
fm_nonconvex_hull(x, ..., format = "sf", method = "fm")
# S3 method for sfg
fm_nonconvex_hull(x, ..., format = "sf", method = "fm")
# S3 method for fm_segm
fm_nonconvex_hull(x, ..., format = "sf", method = "fm")
# S3 method for fm_segm_list
fm_nonconvex_hull(x, ..., format = "sf", method = "fm")

Value

fm_nonconvex_hull() returns an extended object as an sfc polygon object (if format = "sf") or an fm_segm object (if `format = "fm")

fm_extensions() returns a list of sfc objects.

Arguments

x

A spatial object

...

Arguments passed on to the fm_nonconvex_hull() sub-methods

format

character specifying the output format; "sf" (default) or "fm"

method

character specifying the construction method; "fm" (default) or "sf"

convex

numeric vector; How much to extend

concave

numeric vector; The minimum allowed reentrant curvature. Default equal to convex

resolution

integer; The internal computation resolution. A warning will be issued when this needs to be increased for higher accuracy, with the required resolution stated. For method="fm" only.

eps, eps_rel

The polygonal curve simplification tolerances used for simplifying the resulting boundary curve. See fm_simplify_helper() for details. For method="fm" only.

crs

Optional crs object for the resulting polygon. Default is fm_crs(x)

preserveTopology

logical; argument to sf::st_simplify() (for method="sf" only)

dTolerance

If not zero, controls the dTolerance argument to sf::st_simplify(). The default is pmin(convex, concave) / 40, chosen to give approximately 4 or more subsegments per circular quadrant. (for method="sf" only)

Functions

• fm_extensions(): Constructs a potentially nonconvex extension of a spatial object by performing dilation by convex + concave followed by erosion by concave. This is equivalent to dilation by convex followed by closing (dilation + erosion) by concave.

  The ... arguments are passed on to fm_nonconvex_hull_fm() or fm_nonconvex_hull_sf(), depending on the method argument.

• fm_nonconvex_hull_fm(): fmesher method for fm_nonconvex_hull(), which uses the splancs::nndistF() function to compute nearest-neighbour distances.

• fm_nonconvex_hull_sf(): Differs from sf::st_buffer(x, convex) followed by sf::st_concave_hull() (available from GEOS 3.11) in how the amount of allowed concavity is controlled.

INLA compatibility

For mesh and curve creation, the fm_rcdt_2d_inla(), fm_mesh_2d_inla(), and fm_nonconvex_hull_inla() methods will keep the interface syntax used by INLA::inla.mesh.create(), INLA::inla.mesh.2d(), and INLA::inla.nonconvex.hull() functions, respectively, whereas the fm_rcdt_2d(), fm_mesh_2d(), and fm_nonconvex_hull() interfaces may be different, and potentially change in the future.

Details

Morphological dilation by convex, followed by closing by concave, with minimum concave curvature radius concave. If the dilated set has no gaps of width between $$2 \textrm{convex} (\sqrt{1+2\textrm{concave}/\textrm{convex}} - 1) $$ and \(2\textrm{concave}\), then the minimum convex curvature radius is convex.

The implementation is based on the identity $$\textrm{dilation}(a) \& \textrm{closing}(b) = \textrm{dilation}(a+b) \& \textrm{erosion}(b)$$ where all operations are with respect to disks with the specified radii.

When convex, concave, or dTolerance are negative, fm_diameter * abs(...) is used instead.

References

Gonzalez and Woods (1992), Digital Image Processing

See Also

fm_nonconvex_hull_inla()

Examples

inp <- matrix(rnorm(20), 10, 2)
out <- fm_nonconvex_hull(inp, convex = 1, method = "sf")
plot(out)
points(inp, pch = 20)

out <- fm_nonconvex_hull(inp, convex = 1, method = "fm", format = "fm")
lines(out, col = 2, add = TRUE)
if (TRUE) {
  inp <- sf::st_as_sf(as.data.frame(matrix(1:6, 3, 2)), coords = 1:2)
  bnd <- fm_extensions(inp, convex = c(0.75, 2))
  plot(fm_mesh_2d(boundary = bnd, max.edge = c(0.25, 1)), asp = 1)
}