New Non-Iterative Method for Pose Estimation with High Precision

Fig. 2. 200 times experimental results with respect to varying number of points. (a) Rotation matrix error; (b) translation matrix error; (c) focal error; (d) reprojection error

Download full size

Fig. 3. 200 times experimental results with respect to varying noise level. (a) Rotation matrix error; (b) translation matrix error; (c) focal error; (d) reprojection error

Download full size

Fig. 4. Computation time with respect to varying number of point

Download full size

Fig. 5. SIFT points and matching lines

Download full size

Fig. 6. SIFT and reprojection points

Download full size

Fig. 7. 3D reconstruction result of unmanned aerial vehicle dataset

Download full size

Fig. 8. 3D reconstruction result of CastleP30 dataset

Download full size

Table 1. Estimation results of calibration matrix with different datasets

View table

Table 1. Estimation results of calibration matrix with different datasets

Dataset	Number	Estimated calibration matrix	Reference calibration matrix
Unmanned aerial vehicle	50	$\begin{matrix} [\begin{matrix} 3810.73 & 0 & 2457.50 \\ 0 & 3811.18 & 1632.00 \\ 0 & 0 & 1 \end{matrix}] \end{matrix}$	N/A
CastleP30	30	$\begin{matrix} [\begin{matrix} 2761.99 & 0 & 1526.50 \\ 0 & 2766.51 & 1009.50 \\ 0 & 0 & 1 \end{matrix}] \end{matrix}$	$\begin{matrix} [\begin{matrix} 2759.48 & 0 & 1520.69 \\ 0 & 2764.16 & 1006.81 \\ 0 & 0 & 1 \end{matrix}] \end{matrix}$

Table 2. Estimation error of different datasets
View table
Table 2. Estimation error of different datasets
Dataset Number Average reprojectionerror/pixel Average rotationerror /(°) Average translationerror/%
Unmanned aerial vehicle 50 0.43 N/A N/A
CastleP30 30 0.27 0.08 0.11