Skip to contents

Produces a single-channel sg_image from a channel index, a channel or band name, a band wavelength, or one of the registered band operations band_mean, ratio and normalized_difference. No free expressions are accepted. The result records the source bands, wavelengths, operation, value semantics and a transform digest so declarative protocol runs and direct calls stay comparable.

Usage

sg_select_channel(
  image,
  channel = 1L,
  channel_name = NULL,
  wavelength_nm = NULL,
  wavelength_tolerance_nm = 5,
  band_operation = "none",
  band_a_nm = NULL,
  band_b_nm = NULL,
  band_min_nm = NULL,
  band_max_nm = NULL
)

Arguments

image

An sg_image object.

channel

1-based channel index (used when no other selector is given).

channel_name

Channel or band name.

wavelength_nm

Band centre wavelength to select.

wavelength_tolerance_nm

Maximum distance to the nearest band centre.

band_operation

One of "none", "band_mean", "ratio", "normalized_difference".

band_a_nm, band_b_nm

Wavelengths of bands a and b for ratio and normalized_difference.

band_min_nm, band_max_nm

Wavelength interval for band_mean.

Value

A single-channel sg_image.

Details

Band operations require band wavelengths (see new_sg_image() bands or sg_read_envi()) and refuse images whose value_semantics is "unknown" for ratio and normalized_difference, because such ratios are only interpretable for a declared value scale. Non-finite results (division by zero) are replaced by the smallest finite result and counted in provenance$band_selection$non_finite_replaced.

Examples

cube <- array(runif(5 * 5 * 3), dim = c(5, 5, 3))
img <- new_sg_image(cube, channels = c("b450", "b550", "b650"),
                    bands = data.frame(wavelength_nm = c(450, 550, 650)),
                    value_semantics = "reflectance")
sg_select_channel(img, wavelength_nm = 552)$channels
#> [1] "b550"
nd <- sg_select_channel(img, band_operation = "normalized_difference",
                        band_a_nm = 650, band_b_nm = 550)
nd$provenance$band_selection$operation
#> [1] "normalized_difference"