Sie sind auf Seite 1von 8

Susan T.

Williams

GEOG 484

Lesson 3:
Georeferencing Raster Images

Winter 2012

I - A. Georeferencing of statecollege_DRG.tif

Figure 1: Screen capture after completing the georeferencing for statecollege_DRG.tif which shows the
activated layer in the Table of Contents, the Link Table containing the links and Transformation method and
RMS error, and the coordinate readout in Meters.

I - B. Calculating RMS error:


During the course of the lesson, we were asked to calculate the Root Mean Square (RMS) error for Part II
Section B, and also to determine whether or not the RMS error result contained any diagnostic value and, if so,
whether the value was acceptable. We are instructed that the general rule-of-thumb is that the RMS error be
less than or equal to one half the side dimension, in map projection units, of a cell (pixel) (Sloan, 2012).
In order to calculate the RMS error, we need to know the scale and resolution of the map.
In Part II-section B-step 6, we are told that the map in question has a scale of 1:24,000 with a resolution of 250
cells per inch.
th
Therefore, in this case, one-half the side dimension of a cell (pixel) is 1/250 of one inch x (which equals
th
1/500 of one inch).
At a scale of 1:24,000, one inch is equal to 2,000 feet. (24,000 inches / 12 inches per foot = 2,000 feet)
th
So 2,000 feet x 1/500 = 4 feet (2,000 / 500 = 4).
However, the result must be given in map projection units, which are meters in this case.
1 foot = 0.3048 meters, so we multiply 4 feet x 0.3048 = 1.2192 meters.
If we can obtain an RMS error of less than or equal to 1.2192 meters, it will be acceptable in this case. (For
simplicitys sake, we can round it to 1.22 meters.)
Although I did delete and replace control points a couple of times, I am now quite content with an acceptable
RMS value of 0.378.

I - C. Discussion of RMS error for the georeferencing exercise:


Does the RMS error that you ended up with for the statecollege_DRG.tif have any diagnostic
value?
RMS error can only have diagnostic value if specific information is known about the input map - such as scale,
resolution, and the projection, as well as the projection of the target map. Additionally, a minimum number of
control points is required based on the transformation method and their placement is critical to accuracy (ibid).
If this information is known, it becomes possible to calculate a valid RMS error and therefore the RMS error
would have diagnostic value.
In this example with statecollege_DRG.tif, we know the scale and resolution and are therefore able to calculate
the cell size and RMS error. We know the coordinate systems that were utilized and an appropriate number of
control points were specified. Therefore, all critical information was known and the RMS error does therefore
have diagnostic value in this case.

Is the RMSE value acceptable?


I ended up with an RMS value of 0.378, which is considered an acceptable error as it is less than the 1.22
meters previously calculated as being an acceptable threshold.

II - A. Georeferencing statecollege_map.tif:

Figure 2: Screen capture after completing the georeferencing for statecollege_map.tif which shows the
activated layer in the Table of Contents, the Link Table containing the links and Transformation method and
RMS error, and the coordinate readout in Inches.

II - B. Calculating RMS error:


As this is the scanned paper map of the previous DRG image, our RMSE value will be the same as previously
calculated: 1.22 meters.

II - C. Discussion of RMS error for the georeferencing exercise:


Does the RMS error that you ended up with for the statecollege_map.tif have any diagnostic
value?
RMS error can only have diagnostic value if specific information is known about the input map - such as scale,
resolution, and the projection, as well as the projection of the target map. Additionally, a minimum number of
control points is required based on the transformation method and their placement is critical to accuracy (ibid).
If this information is known, it becomes possible to calculate a valid RMS error and therefore the RMS error
would have diagnostic value.
In this example with statecollege_map.tif, we have some issues that complicate the RMS error. In spite of
knowing the coordinate systems of both input and target layers as well as having four control points, as we did
with our first image, this particular map ended up with an RMS error of 7.2 meters, which is quite far out of the
acceptable range of 1.22 meters. The control points were in different areas of the map, spaced fairly far apart,
but they are not all evenly distributed and on the edges of the map as ideal control points should be. However,
it is the rather large and formidable wrinkle in the upper portion of the map that most likely contributes the
greatest to this unacceptably high RMS value. Although the RMS error is unacceptable, it does have some
minor diagnostic value in this situation, as an appropriate number of control points were used and in the same
coordinate system as the TIFF image.
I like to think of it as being akin to a medical test to check ones levels of, say, iron in the blood. The labwork
might come back showing that the blood levels of iron are not within normal ranges, but this does not
necessarily mean that the labwork itself is inaccurate! Although the level of iron may not be within acceptable
ranges, the test does have diagnostic value as it helps indicate that something is indeed abnormal and the
patient has anemia. Following this analogy, I feel it fair to say that the RMS error in this instance may not be
within acceptable ranges but still holds some diagnostic value in that it tells us the accuracy of this map is not
very reliable and so the image could possibly be used for basic information but it would be better to utilize a
different source if any great accuracy is needed.

III - A. Georeferencing for the vertical air photo of State College:

Figure 3: Screen capture after completing the georeferencing for the vertical air photo of State College which
shows the activated layer in the Table of Contents, the Link Table containing the links and Transformation
method and RMS error, and the coordinate readout in Meters.

III - B. Calculating RMS error:


Not applicable to this exercise: please refer to IIIC below.

III - C. Discussion of RMS error for the georeferencing exercise:


Does the RMS error that you ended up with for the vertical air photo of State College have any
diagnostic value?
In this third example, the RMS error is neither acceptable nor diagnostic.
As discussed previously, a RMS error can only have diagnostic value if specific information is known about the
input and target maps, and if there is an appropriate number of well-placed control points. The aerial photo of
State College is not orthorectified, meaning that it has not been corrected for scale variations that normally
occur when taking aerial photos due to differing elevations of locations within the photo. Therefore, the scale in
this photo is not consistent and any measurements of distance or direction would not be accurate (Okeke,
2006).
Additionally, the aerial photo is not associated with a defined coordinate system and we therefore do not have
the necessary information to calculate cell size and RMSE.
As the first 5 control points were added, I had an RMS error of about 15. The RMSE dropped a bit to 12.8 after
adding three extra control points. After the sixth control point was added, I was given an additional choice in
nd
the Transformation dropdown list: 2 Order Polynomial. Once this new option was selected, my RMS error
nd
dropped significantly to 2.8. The 2 Order Polynomial Transformation utilized warping, which resulted in some
of the previously straight edges being deformed and curved slightly (Sloan, 2012).
Bolstad (2008) warns us that utilizing a higher-order polynomial transformation will also introduce more error, as
it allows more warping to fit the surface around the control points, and should therefore be used with caution.
These RMSE values are really not to be taken seriously in this particular instance. If we return to the analogy of
the labwork to check levels of iron in ones blood, this particular situation would be akin to poor lab techniques.
If the lab technician who analyzed the blood did not prepare the sample properly for analysis, for example, the
results of the test would most likely be skewed. This photo has not been properly orthorectified for analysis and
will therefore give skewed results, thus having no diagnostic value.

D. List of factors possibly affecting RMS error result:


1. Control point coordinates may be digitized poorly or incorrectly due to any operator error (ie fatigue or
lack of attention, poor eyesight or misidentification, a shaky hand or a flat-out blunder)
2. Poorly chosen reference point locations (ie- control point is a wide or curved feature, or is obscured by
other features such as trees)
3. Poorly distributed control points (ie not evenly spaced, not around edges of map)
4. Too few control points for the type of Transformation to be used
5. Damaged or wrinkled source data (ie the scanned paper map contains folds or tears that distort the
image)
6. Differences between source and target control point locations (ie placing the point on the wrong side
of the road, or perhaps basing the point on a feature that has since moved)
7. Errors made when copying and pasting or typing in coordinates to the Link Table
(Sloan, 2012 and Bolstad, 2008)

REFERENCES
Bolstad, Paul. (2008) GIS Fundamentals: A First Text on Geographic Information Systems, Third Edition. Elder
Press, White Bear Lake, MN.
Okeke, Dr. Francis I. Review of Digital Image Orthorectification Techniques. GIS Development: Asia Pacific.
July 2006. Accessed January 29, 2012 from
http://geospatialworld.net/index.php?option=com_content&view=article&id=20477&Itemid=2063
Sloan, J. (1999-2012). GIS Database Development, Lesson 3. The Pennsylvania State University World
Campus Certificate Program in GIS. Accessed January 27, 2012 from https://www.eeducation.psu.edu/geog484/l3.html
All images from screen captures produced with ArcMap v10.0 by Environmental Systems Research Institute
(ESRI).

This document is published in fulfillment of an assignment by a student enrolled in an educational offering of the
Pennsylvania State University. The student, named above, retains all rights to the document and responsibility for its
accuracy and originality.

Das könnte Ihnen auch gefallen