14.11.2013 Views

Fernerkundung I (Digitale Bildverarbeitung) - Friedrich-Schiller ...

Fernerkundung I (Digitale Bildverarbeitung) - Friedrich-Schiller ...

Fernerkundung I (Digitale Bildverarbeitung) - Friedrich-Schiller ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Determining the Accuracy of the Polynomial Fit GCP Works 9 be determined by selecting<br />

Model Coefficients from the Reports menu on the GCP Selection and Editing panel.<br />

Determining the Accuracy of the Polynomial Fit:<br />

In order to determine the accuracy of the derived coefficients of the GCPs, examine the<br />

results of the least squares regression of the initial GCPs:<br />

• The GCP scatter plot report, which can be activated by selecting GCP Scatter Plot from<br />

the Reports menu on the GCP Selection and Editing panel, shows the X and Y residual<br />

errors for each GCP point on a crosshair graph.<br />

• The RMS error for each GCP is reported in the lists of Accepted and Check GCPs on the<br />

GCP Selection and Editing panel. A total RMS error is also reported on this panel.<br />

The RMS error is calculated in pixel units by these equations:<br />

RMS error = ( ( x1 - xorg ) 2 + ( y1 - yorg ) 2 ) 1/2<br />

x1 = computed row coordinate in the uncorrected image.<br />

y1 = computed column coordinate in the uncorrected image.<br />

xorg = original row coordinate of the GCP in the image.<br />

yorg = original column coordinate of the GCP in the image.<br />

By computing the RMS errors for all of the GCPs, it is possible to see which GCPs exhibit<br />

the greatest error and to sum the RMS errors. If a given set of control points produce a<br />

total RMS error that exceeds your acceptable limit, you should consider deleting the<br />

GCPs that have the greatest errors. An error of less than the dimension of one pixel is<br />

suggested as an acceptable limit.<br />

The Image Fit Report is also available to help you assess the fit of the regression model.<br />

This report can be activated by selecting Image Fit from the Reports menu on the GCP<br />

Selection and Editing panel. It shows a graphical representation of the outline of the<br />

georeferenced image area, uncorrected image area, mosaic cut line, and GCP points. The<br />

uncorrected image outline is transformed according to the current GCP model. This<br />

preview of how the uncorrected image would map onto the georeferenced data set with<br />

the current set of GCPs allows the registration to be visually assessed before it is<br />

performed.<br />

Thin Plate Spline:<br />

Thin Plate Splines (TPSs) provide an attractive alternative to the traditional polynomials.<br />

Thin Plate Splines are also global (i.e., all GCPs are used simultaneously to derive the<br />

transformation), but the derived functions have minimum curvature between control<br />

points and become almost linear at great distances from the GCPs. The influence of an<br />

individual GCP is localized and diminishes rapidly the further away from the points. The<br />

TPS functions interpolate the values at all ground control points, within the numerical<br />

round-off error limits, and therefore a GCP can always be added in an area where the<br />

transformation is not satisfactory. The main disadvantage of TPSs is that, to represent a<br />

warping transformation accurately, they should be constrained at all extreme points of<br />

the warping function. This is not a problem in smoothly varying transformations, such as<br />

the change of coordinate systems (if the exact transformation functions are<br />

not known). However, when using TPSs to georegister a photograph in rough terrain, it<br />

may be necessary to acquire hundreds of GCPs, since there should be a point at every<br />

extreme of terrain (peak or valley bottom), and along breaklines. The derivation of TPS<br />

transformation parameters involves solving an equation system with a square (N+3) by<br />

(N+3) matrix, where N is the number of GCPs. For a very large N it is a time consuming<br />

task. Moreover, the evaluation of TPS functions at every image pixel requires calculation<br />

of N values of natural logarithms, and for large N this calculation may be prohibitively

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!