View Full Version : Number of control points in DLT

Bing Yu
03-10-1998, 12:31 AM
When I was at the University of Iowa, we used up to 68 control points for the
calibrations in DLT. I prefer using more than 16 control points for two
reasons: (a) using more control points allows me to eliminate some
inacurrately digitized or incorrect defined points; and (b) using more control
points allows me to do extrapolations if I have to.

As Dr. Brian Davis pointed out, we need to eliminate those inaccurately
digitized or incorrectly defined control points. My own DLT program package
can do this automatically or let users to do this themselves. Most of time,
this "optimization" of the calibration can reduce overall calibration error by
1-2 mm. This improvement sometimes is indeed greater than that by increasing
the number of calibration points.

I found that using more control points significantly improves the accuracy of
extrapolation. Thoretically, we should avoid extrapolation as much as we can.
In reality, we have to more or less do some extrapolations. With the number
of control points between 60-68, I did not see significant decrease in the
accuracy in the estimated 3-D coordinates even the point was over 30 cm
outside the calibration volume. I do feel that PEAK PERFORMANCE 24 contropl
point calibration frame is not sufficient for extrapolation although it may
provide reasonable interpolation accuracy.

Overall, I agree with Rick. We should use more than 16 control points for DLT
calibration. I don't know the exact number of control points that should be
used as minimum, but I think that using more control points definitely has
some advantages and digitizing 50 or more control points does cost signicantly
more time than that for 16 control points.

Bing Yu, Ph.D.
Assistant Professor
Division of Physical Therapy
The University of North Carolina at Chapel Hill