Open the Geographic Calculator.
Click on the Point Database Conversions job to open the interface.
Load
a point database file by clicking the Load Data button below the grid
table on the left.
On the dialog that opens, select the radio button Use File, and click
the Browse button to locate the file you will open. The
file to open is in teh folder for the Getting Started Guide (C:\Program
Files\Blue Marble Geo\Geographic Calculator 7.0\Getting Started Guide)
named Best-Fit_Input.xls.
When
you open the file, you will be asked if the file contains a Header. Select
Yes. This
will use the first row of the spreadsheet to name the columns in the table.
Once loaded,
the table should look like this:
In
the Operation area at the bottom of the interface, select the radio button
for Best Fit. When
you do so, the Interface will change to display a new fitted coordinate
system definition on the left hand coordinate system.
The local coordinates in this exercise are in Meters, so you will not need to change the units.
To
assign a name to the fitted system you will create, double click the blue
box that says "New Fitted Coordinate System". In
the dialog that pops up, go to the Identification
tab and change the Name to "GSG Sample System". Click
OK.
Your
control coordinates are in UTM Zone 15 North, WGS84, Meters. To
select the output coordinate system, double-click on the blue box labeled
System in the Coordinate System area at the bottom right of Point Database
job. By
default, the Calculator starts with the geodetic coordinate system of
WGS84 selected. This
will launch the Select Coordinate System dialog.
To select the Universal Transverse Mercator Zone 15 North, in the tree
view on the left, browse under Projected>UTM>WGS84,
in the coordinate system list on the right, select Zone
15N. Then
click OK.
Click
on the Conversion Settings button.
This will display a form to specify which columns of data in the point
database are to be used to compute the transformation. Select the Best Fit tab to open the Best Fit-specific
settings:
The local coordinates from the system being registered are in the Local North and Local East columns. Specify those columns in the Original area of the Input side.
The Transform coordinates are the control coordinates that the Fit will be based on. Set these columns to the UTM North and UTM East columns.
The Point Use Flag is a column that will act as a switch to either include or exclude particular control pairs from the Best Fit model. Click the Add button to add this column and name it "Point Flag". When you do so, a column will be added containing a "1" on every row. The 1 will include that control pair in the Fit model. We will come back to this later.
Leave the Point Name set to (none) because we do not have a column containing unique names for the control points.
Select the Math Transform Affine.
The remaining parameters should be assigned to empty columns that you add to the point database grid. They can automatically be added by clicking the Add button next to each output parameter and specifying a new column be added after the last column in the list. It is a good idea to simply name each new column with the names shown for each drop down.
In the Flag Errors Greater Than area, set the field to 3 Times RMSE (Root Mean Square Error).
The
Settings should now look like this:
Click OK to initialize these
settings.
Click
on the Calculate button to compute
the transformation. If the processing was successful, you will see error
values in the columns you added for North Error, East Error, North Revers
Error, and East Reverse Error. Also,
the error plot tool button and menu item will be enabled. The
table should look like this:
Once
a Best Fit has been calculated, you can view a plot displaying the relative
errors to assist in finding errors. To
View the Error Plot, click on the Error Plot button on the Toolbar at
the top.
In
the plot, the locations of all of the points in the point database are
shown, with a line drawn from each signifying the direction and relative
magnitude of the point’s residual error (the difference between its file-specified
location and its computed location from the transformation). The tabs
at the top of the Error Plot allow you to alternate between the display
of the residual errors from the forward transformation and those of the
reverse transformation.
Points which were used to compute the coordinate system transformation are shown with a circle and are normally drawn in green. Those points which were excluded are drawn in black. A point whose residual error has a magnitude exceeding the criteria specified by the Best-Fit conversion settings (in this case three times the root mean square of the residual errors) are drawn in red. The RMSE indicates the overall error in the units specified for the coordinate systems, in this case Meters.
Using
the Error Plot it is fairly obvious that one of the control points is
questionable. Click
on the red point with the large vector coming off of it to see the ID
or the point and the errors associated with it.
We can see that there is an X error of over 1 meter for point R21 (the
point in Row 21).
To remove that control point from the model, click back to the Point Database Job by clicking the tab for it above the error plot.
In
the Table, scroll down to row 21 and change the value of the Point Flag
column from "1" to "0". When
the fit is computed, the 0 will remove that control point from the model.
Then click
Calculate to recalculate the Best
fit model. You
will see the error values all change to reflect the new model.
Click
the Error Plot button to see the new Error Plot for the model. The effect
of excluding the points should be readily apparent in the new plot. The
removed point will be displayed in black. The
RMSE for the whole model will also have decreased significantly, indicating
a better fit of the points. The
clustered points now fit much better to the computed transformation (which
is also reflected in a lower root-mean-square residual error), while the
excluded points have greater errors.
NOTE: While excluding points is useful to remove suspect points
from the transformation, care must be taken to not exclude so many points
that the transformation may not be accurate or reliable throughout the
area of interest.
The
fit parameters are automatically entered into the Geodetic Data Source's
Custom.xml file. However,
to make the fitted system a permanent addition to your Geographic Calculator,
you will need to save the change. To
save the change, go to Options>Coordinate
System Definitions.
To
view your fitted system, in the tree view on the left, select Fitted;
You will see your new system in blue on the right. The
blue text of the name indicates that this entry is in the Custom.xml
file and is editable by the user.
To save the new system, click the Save button on the right.