As stated previously, orthorectification is the process of removing geometric errors inherent within photography and imagery. The variables contributing to geometric errors include, but are not limited to:
- Camera and sensor orientation
- Systematic error associated with the camera or sensor
- Topographic relief displacement
- Earth curvature
By performing block triangulation or single frame resection, the parameters associated with camera and sensor orientation are defined. Utilizing least squares adjustment techniques during block triangulation minimizes the errors associated with camera or sensor instability. Additionally, the use of self-calibrating bundle adjustment (SCBA) techniques along with Additional Parameter (AP) modeling accounts for the systematic errors associated with camera interior geometry. The effects of the Earth’s curvature are significant if a large photo block or satellite imagery is involved. They are accounted for during the block triangulation procedure by setting the relevant option. The effects of topographic relief displacement are accounted for by utilizing a DEM during the orthorectification procedure.
The orthorectification process takes the raw digital imagery and applies a DEM and triangulation results to create an orthorectified image. Once an orthorectified image is created, each pixel within the image possesses geometric fidelity. Thus, measurements taken off an orthorectified image represent the corresponding measurements as if they were taken on the Earth’s surface (see the figure below).
An image or photograph with an orthographic projection is one for which every point looks as if an observer were looking straight down at it, along a line of sight that is orthogonal (perpendicular) to the Earth. The resulting orthorectified image is known as a digital orthoimage (see the figure below).
Relief displacement is corrected by taking each pixel of a DEM and finding the equivalent position in the satellite or aerial image. A brightness value is determined for this location based on resampling of the surrounding pixels. The brightness value, elevation, and exterior orientation information are used to calculate the equivalent location in the orthoimage file.
Digital Orthophoto—Finding Gray Values
P = ground point
P1 = image point
O = perspective center (origin)
X,Z = ground coordinates (in DTM file)
f = focal length
In contrast to conventional rectification techniques, orthorectification relies on the digital elevation data, unless the terrain is flat. Various sources of elevation data exist, such as the USGS DEM and a DEM automatically created from stereo image pairs. They are subject to data uncertainty, due in part to the generalization or imperfections in the creation process. The quality of the digital orthoimage is significantly affected by this uncertainty. For different image data, different accuracy levels of DEMs are required to limit the uncertainty-related errors within a controlled limit. While the near-vertical viewing SPOT scene can use very coarse DEMs, images with large incidence angles need better elevation data such as USGS level-1 DEMs. For aerial photographs with a scale larger than 1:60000, elevation data accurate to 1 meter is recommended. The 1 meter accuracy reflects the accuracy of the Z coordinates in the DEM, not the DEM resolution or posting.
Detailed discussion of DEM requirements for orthorectification can be found in Yang and Williams (Yang and Williams, 1997).
Resampling methods used are nearest neighbor, bilinear interpolation, and cubic convolution. Generally, when the cell sizes of orthoimage pixels are selected, they should be similar or larger than the cell sizes of the original image. For example, if the image was scanned at 25 microns (1016 dpi) producing an image of 9K × 9K pixels, one pixel would represent 0.025 mm on the image. Assuming that the image scale of this photo is 1:40000, then the cell size on the ground is about 1 m. For the orthoimage, it is appropriate to choose a pixel spacing of 1 m or larger. Choosing a smaller pixel size oversamples the original image.
For information, see Scanning Resolutions Table.
For SPOT Pan images, a cell size of 10 meters is appropriate. Any further enlargement from the original scene to the orthophoto does not improve the image detail. For IRS-1C images, a cell size of 6 meters is appropriate.