There may be occasions when the image position of candidate landmarks cannot be considered useful for positioning. For example, rolling terrain or other factors may make it impossible to constrain the pose of the camera in a consistent orientation. In these cases, one might wish to reduce to zero. Figure 6.5 demonstrates the accuracy of position estimation when the image-position parameter, is set to zero, as the camera scale parameter, is varied. The figure effectively plots the surface depicted in Figure 6.4 in the limit as approaches 0. As the plot indicates, purely appearance-based pose estimation is very effective for a wide range of parameterisations.
Figure 6.5: Appearance-based estimation error for Scene I. The plot depicts
the mean estimation error as a function of , the camera scale
parameter when .
For a more specific look at how well the method performs when , Figure 6.6 plots the set of twenty test cases for . The mean estimation error is 0.17cm.
Figure 6.6: Appearance-based estimation results for Scene I. The plot depicts
the set of pose estimates for the twenty test cases in Scene I with
and . The mean estimation error is 0.17cm.