The purpose of this lesson was to gain
experience in georeferencing raster images. The first portion of this exercise
involved georeferencing the State College, PA 7.5 minute topographic quadrangle.
This was conducted by entering the UTM coordinates (NAD27) of four known control
points. The accuracy of a georeferenced image is measured as RMS error. An
acceptable RMS error should equal less than one-half of a pixel in digitized map
units (i.e., meters in the case of UTM) (Sloan 2007). In the case of a DRG, the
resolution is known to be 250 cells per inch. The original scale of the
quadrangle can be determined by reading the information presented in its collar:
1:24,000.
An acceptable RMS error can be calculated
by dividing the scale (1:24,000) by the number of pixels in each inch:
24,000/250 = 96
The next step is to divide this (96) by
the number of inches in one foot:
96/12 = 8
The last step is to convert this number to
meters (there are 0.3048 meters in 1 foot):
8 * 0.3048 = 2.4384 m/pixel
The acceptable RMS error, would therefore
be 2.436 * 0.5 or 1.22 meters. The RMS
error generated while georeferencing the State College, PA topographic quad was
1.12 meters and was within the range of an acceptable RMS error.
Part 1:
The RMS error generated while georeferencing the
statecollege_DRG.tif is of diagnostic value. This image was based on a Digital
Raster Graphic (DRG) that has simply lost its georeferencing. The scale and
resolution were known and RMS error could be calculated. As previously
mentioned, the image had been previously georeferenced and was free of
distortion. An acceptable RMS error for a 1:24,000 DRG is 1.22 meters (see
above). An RMS error of 1.12 meters was generated while attempting to
georeference the DRG. This result is within the range of acceptable RMS error.
Part 2:
The RMS error generated while georeferencing the
statecollege_map.tif is of diagnostic value. As mentioned above, the acceptable
RMS error for a 7.5 minute quadrangle is 1.22 meters. Even though the
statecollege_map.tif is a scanned portion of the quadrangle, the acceptable RMS
error is still the same. The RMS error of the statecollege_map.tif is 7.27
meters, well above the acceptable limit. This high RMS error is a result of
multiple folds and generally poor condition of the scanned map. Even though the
four control points are located below the primary crease on the map, the overall
condition of the map resulted in the high RMS error. The four control points
were also poorly placed and should have been located further apart and closer to
the edges of the map. When the statecollege_map.tif is overlaid on
statecollege_DRG.tif, the poor alignment can be seen along the boundary
(Appendix A).
Part 3:
The RMS error generated for the vertical unorthorectified
aerial photo of State College does not have any diagnostic value. Initially, the
image was georeferenced based on five control points and a 1st Order Polynomial
(Affine) transformation. This resulted in a seemingly high RMS Error (approx.
17.0 units). Additional control points were added and a 2nd Order Polynomial
transformation was conducted. This resulted in a relatively lower RMS error of
1.01 units. The 2nd order transformation warped the raster image to better fit
the control points. Because the projection is unknown for the aerial photograph,
the scale cannot be determined, making it impossible to calculate an acceptable
RMS error. The RMS error of 1.01 is smaller than the previous 17.0 but because
the units are not known, this number is of little use. Furthermore, because the
ground surface elevation has not been corrected for (as in the case of
orthophotos), the scale is not consistent across the entire image.
