Skip to contents

Plot the Learned Correction Surface

plot_correction_surface.Rd

Visualizes the spatial correction field (dx, dy) learned by a PAI model.

Usage

plot_correction_surface(pai_model, gcp_data, n_grid = 100)

Arguments

pai_model

A trained pai_model object returned by train_pai_model().

gcp_data

The sf object of homologous points that was used for training the model.

n_grid

The resolution of the interpolation grid used to create the smooth surface. Higher values create a more detailed plot but take longer to compute. Defaults to 100.

Value

A patchwork object containing two ggplot plots, one for dx and one for dy.

Details

This function serves as a key diagnostic tool for understanding the behavior of a trained pai_model. It creates two raster plots: one for the dx (East-West) corrections and one for the dy (North-South) corrections.

The color intensity on the plots reveals the magnitude of the correction at any given location. Contour lines show the gradient of the change, and black crosses mark the location of the original Ground Control Points (GCPs), showing where the model had direct information to learn from.

By examining these surfaces, users can:

  • Understand the spatial nature of the distortion their model has learned.

  • Identify areas of high vs. low correction.

  • Spot potential issues like extreme corrections or unusual artifacts, especially at the edges of the data where the model is extrapolating.

Examples

if (FALSE) { # \dontrun{
# This example demonstrates how to train a model and then visualize
# what it has learned.

# --- 1. Load required data from the package ---
data(gcps)

# --- 2. Train a model to visualize ---
# A GAM (Generalized Additive Model) is a great choice for this
# because it naturally produces smooth surfaces that are easy to interpret.
pai_model_gam <- train_pai_model(gcps, method = "gam")

# --- 3. Generate and display the plot ---
# This creates the two-panel plot showing the dx and dy correction fields.
correction_plot <- plot_correction_surface(
  pai_model = pai_model_gam,
  gcp_data = gcps,
  n_grid = 75 # Using a slightly coarser grid for a quick example
)

# To display the plot in an interactive session:
print(correction_plot)
} # }