Flags:

-c

Use current region settings in target location (def.=calculate smallest area)

-a

Rectify all raster maps in group

-t

Use thin plate spline

--help

Print usage summary

--verbose

Verbose module output

--quiet

Quiet module output

--ui

Force launching GUI dialog

Parameters:

group=name [required]

Name of input imagery group

input=name[,name,...]

Name of input raster map(s)

extension=string [required]

Output raster map(s) suffix

order=integer [required]

Rectification polynomial order (1-3)
Options: 1-3
Default: 1

resolution=float

Target resolution (ignored if -c flag used)

memory=memory in MB

Maximum memory to be used (in MB)
Cache size for raster rows
Default: 300

method=string

Interpolation method to use
Options: nearest, linear, cubic, lanczos, linear_f, cubic_f, lanczos_f
Default: nearest

Coordinate transformation

The desired order of transformation (1, 2, or 3) is selected with the order option. The program will calculate the RMSE and check the required number of points.

Linear affine transformation (1st order transformation)

x’ = ax + by + c
y’ = Ax + By + C The a, b, c, A, B, C are determined by least squares regression based on the control points entered. This transformation applies scaling, translation and rotation. It is NOT a general purpose rubber-sheeting like TPS, nor is it ortho-photo rectification using a DEM, not second order polynomial, etc. It can be used if (1) you have geometrically correct images, and (2) the terrain or camera distortion effect can be ignored.

Polynomial Transformation Matrix (2nd, 3d order transformation)

i.rectify uses a first, second, or third order transformation matrix to calculate the registration coefficients. The number of control points required for a selected order of transformation (represented by n) is
((n + 1) * (n + 2) / 2) or 3, 6, and 10 respectively. It is strongly recommended that one or more additional points be identified to allow for an overly-determined transformation calculation which will generate the Root Mean Square (RMS) error values for each included point. The RMS error values for all the included control points are immediately recalculated when the user selects a different transformation order from the menu bar. The polynomial equations are performed using a modified Gaussian elimination method.

Thin plate spline (TPS) transformation

TPS transformation is selected with the -t flag. This method of coordinate transformation is recommended for satellite imagery where hundreds or thousands of GCPs are included, and for historical printed or scanned maps with unknown georeferencing and/or known localized distortions.

TPS combines a linear affine transformation with individual transformation coefficients for each GCP, using the radial basis kernel function with the distance dist between any two points:
dist2 * log(dist) As a consequence, localized distortions can be removed with TPS transformation. For example, scan line sensors will have due to the changing viewing angle larger distortions towards the end points of the scan line than at the center of the scan line. Even higher order polynomial transformations are not able to remove these locally different distortions, but TPS transformation can. For best results, TPS requires an even and, for localized distortions, dense spacing of GCPs.

Resampling method

The rectified data is resampled with one of seven different methods: nearest, bilinear, cubic, lanczos, bilinear_f, cubic_f, or lanczos_f.

The method=nearest method, which performs a nearest neighbor assignment, is the fastest of the resampling methods. It is primarily used for categorical data such as a land use classification, since it will not change the values of the data cells. The method=bilinear method determines the new value of the cell based on a weighted distance average of the 4 surrounding cells in the input map. The method=cubic method determines the new value of the cell based on a weighted distance average of the 16 surrounding cells in the input map. The method=lanczos method determines the new value of the cell based on a weighted distance average of the 25 surrounding cells in the input map.

The bilinear, cubic and lanczos interpolation methods are most appropriate for continuous data and cause some smoothing. These options should not be used with categorical data, since the cell values will be altered.

In the bilinear, cubic and lanczos methods, if any of the surrounding cells used to interpolate the new cell value are NULL, the resulting cell will be NULL, even if the nearest cell is not NULL. This will cause some thinning along NULL borders, such as the coasts of land areas in a DEM. The bilinear_f, cubic_f and lanczos_f interpolation methods can be used if thinning along NULL edges is not desired. These methods "fall back" to simpler interpolation methods along NULL borders. That is, from lanczos to cubic to bilinear to nearest.

If nearest neighbor assignment is used, the output map has the same raster format as the input map. If any of the other interpolations is used, the output map is written as floating point.