next up previous contents
Next: Normalization with #tex2html_wrap_inline2635# Up: Least-Squares Fitting Based on Previous: Least-Squares Fitting Based on

Normalization with A+C=1

 

Since the trace A+C can never be zero for an ellipse, the arbitrary scale factor in the coefficients of the conic equation can be naturally removed by the normalization A+C=1. This normalization has been used by many researchers [17, 21]. All ellipse can then be described by a 5-vector

displaymath2593

and the system equation tex2html_wrap_inline2579 becomes:

  equation73

where

displaymath2594

displaymath2595

Given n points, we have the following vector equation:

displaymath2596

where

eqnarray83

The function to minimize becomes

displaymath2597

Obtaining its partial derivative with respect to tex2html_wrap_inline2527 and setting it to zero yield:

displaymath2598

The solution is readily given by

  equation104

This method is known as pseudo inverse technique.



Zhengyou Zhang
Thu Feb 8 11:42:20 MET 1996