Assess affine transformation quality from matched GCPs

Fit a first-order affine transformation from matched GCPs using least squares and return per-GCP residuals. Use this to identify poorly digitized points before the expensive warp step.

Usage

hsi_check_gcp(x, verbose = FALSE)

Arguments

x

A data.frame or tibble of matched GCPs from hsi_match_gcp(). Must contain columns source_x, source_y, target_x, target_y.

verbose

Logical. Print GCP count and RMSE to console. Default FALSE.

Value

A named list containing:

residuals

A tibble with all input columns plus residual_x, residual_y, and residual_total in target pixels.

rmse

Numeric. Root mean square error in target pixels.

n_gcps

Integer. Number of GCPs used.

Details

The affine model (6 parameters) handles translation, rotation, independent X/Y scaling, and shear. With N GCPs, residual assessment has N - 3 degrees of freedom. Residuals are in target pixel units. An RMSE above 5 pixels triggers a warning.

See also

hsi_match_gcp() for preparing input, hsi_coregister() for applying the warp.

Other HSI Co-registration: hsi_coregister(), hsi_match_gcp()

Examples

if (FALSE) { # \dontrun{
matched <- hsi_match_gcp(swir_gcps, vnir_gcps)
x_check <- hsi_check_gcp(matched)

# Inspect worst GCPs
x_check$residuals |>
  dplyr::arrange(dplyr::desc(residual_total))

# Remove outliers and re-check
cleaned <- matched |> dplyr::filter(!gcp_id %in% c(5, 12))
hsi_check_gcp(cleaned)
} # }