Skip to contents

Apply a step-by-step Vis-NIR / MIR preprocessing pipeline

Source: R/spectra-preprocess.R
apply_spectral_preprocessing.Rd

Composes the canonical soil-spectroscopy sequence, each step optional and applied in this fixed scientific order: reflectance \(\to\) (absorbance) \(\to\) (Savitzky-Golay smoothing) \(\to\) (Savitzky-Golay 1st or 2nd derivative). Each Savitzky-Golay pass trims (window - 1) / 2 columns from each edge; the wavelength axis is trimmed to match and returned so callers can plot the treated spectrum on the correct axis.

Usage

apply_spectral_preprocessing(
  X,
  wavelengths = NULL,
  absorbance = FALSE,
  sg_smooth = FALSE,
  sg_derivative = 0L,
  window = 11L,
  poly = 2L
)

Arguments

X

Numeric matrix (rows = samples/horizons, columns = wavelengths) or a numeric vector (treated as one sample).

wavelengths

Optional numeric wavelength axis. Defaults to the numeric part of colnames(X), else 1:ncol(X).

absorbance

Logical; apply \(A = \log_{10}(1 / R)\).

sg_smooth

Logical; apply Savitzky-Golay smoothing (m = 0).

sg_derivative

Integer 0, 1 or 2; Savitzky-Golay derivative order (0 = none).

window

Odd Savitzky-Golay window (default 11); coerced to a valid odd value in [3, ncol).

poly

Savitzky-Golay polynomial order (default 2); clamped to [1, window - 1].

Value

A list with X (the treated numeric matrix, wavelength column names trimmed to match), wavelengths (numeric) and steps (an ordered character vector describing the transforms actually applied, starting with "Reflectance").

Details

The transform is robust: reflectance that looks like a percentage (maximum > 1.5) is rescaled to a 0–1 fraction before the absorbance log, values are clamped away from zero to avoid log(0), and a Savitzky-Golay step that cannot fit the requested window into the available wavelengths is skipped (recorded in steps) rather than erroring.

See also

preprocess_spectra

Examples

X <- matrix(seq(0.1, 0.5, length.out = 3 * 60), nrow = 3, byrow = TRUE)
colnames(X) <- seq(400, 2400, length.out = 60)
res <- apply_spectral_preprocessing(X, absorbance = TRUE,
                                    sg_smooth = TRUE, sg_derivative = 1L)
res$steps          # ordered treatment labels
#> [1] "Reflectance"                   "Absorbance (log 1/R)"         
#> [3] "SG smoothing (w=11, p=2)"      "SG 1st derivative (w=11, p=2)"
dim(res$X)         # columns trimmed by the two SG passes
#> [1]  3 40