Inertial Navigation Aided Image Feature Matching Method

Fig. 7. Source images. (a) First image of experimental table; (b) second image of experimental table; (c) first image of laboratory; (d) second image of laboratory; (e) first image of office; (f) second image of office

Download full size

Fig. 8. Brute-force match results of three sets of images

Download full size

Fig. 9. Possible areas. (a) Four feature points selected in the first image of experimental table; (b) possible areas corresponding to the feature points in the first image of experimental table; (c) four feature points selected in the first image of laboratory; (d) possible areas corresponding to the feature points in the first image of laboratory; (e) four feature points selected in the first image of office; (f) possible areas corresponding to the feature points in the first image of office

Download full size

Fig. 10. Match results of three sets of images after adding constraints

Download full size

Fig. 11. Match results of three sets of images after removing outliers by RANSAC

Download full size

Table 1. Rotation matrix

View table

Table 1. Rotation matrix

Scene	Rotation matrix
Experimentaltable	$\begin{matrix} [\begin{matrix} 0.948665 & - 0.005961 & 0.316228 \\ 0.008123 & 0.999952 & - 0.005519 \\ - 0.316180 & 0.007804 & 0.948667 \end{matrix}] \end{matrix}$
Laboratory	$\begin{matrix} [\begin{matrix} 0.998534 & 0.004594 & - 0.053926 \\ 0.002084 & 0.992389 & 0.123124 \\ 0.054082 & - 0.123056 & 0.990925 \end{matrix}] \end{matrix}$
Office	$\begin{matrix} [\begin{matrix} 0.996481 & - 0.048260 & - 0.068532 \\ 0.061619 & 0.976050 & 0.208639 \\ 0.056821 & - 0.212128 & 0.975589 \end{matrix}] \end{matrix}$

Table 2. Translation vector
View table
Table 2. Translation vector
Scene Translation vector /mm
Experimentaltable $\begin{matrix} {[\begin{matrix} 13.473121 & 0.004500 & 2.186387 \end{matrix}]}^{T} \end{matrix}$
Laboratory $\begin{matrix} {[\begin{matrix} 9.200683 & - 4.771147 & 0.862319 \end{matrix}]}^{T} \end{matrix}$
Office $\begin{matrix} {[\begin{matrix} 4.823290 & 1.268367 & - 0.356531 \end{matrix}]}^{T} \end{matrix}$

Table 3. Fundamental matrix

View table

Table 3. Fundamental matrix

Scene	Fundamental matrix
Experimental table	$\begin{matrix} [\begin{matrix} - 1.565037 \times 10^{- 9} & - 1.791871 \times 10^{- 7} & 1.964372 \times 10^{- 4} \\ 5.191469 \times 10^{- 7} & - 9.682601 \times 10^{- 9} & - 0.004100 \\ - 5.179737 \times 10^{- 4} & 0.004090 & - 0.100741 \end{matrix}] \end{matrix}$
Laboratory	$\begin{matrix} [\begin{matrix} - 2.130416 \times 10^{- 8} & - 2.201791 \times 10^{- 8} & - 1.334185 \times 10^{- 3} \\ 2.979019 \times 10^{- 8} & 9.308407 \times 10^{- 8} & - 2.758886 \times 10^{- 3} \\ 1.364791 \times 10^{- 3} & 2.548693 \times 10^{- 3} & 9.785125 \times 10^{- 1} \end{matrix}] \end{matrix}$
Office	$\begin{matrix} [\begin{matrix} 7.710623 \times 10^{- 9} & 6.469725 \times 10^{- 9} & 3.591123 \times 10^{- 4} \\ - 5.158159 \times 10^{- 8} & 8.523534 \times 10^{- 8} & - 1.365610 \times 10^{- 3} \\ - 2.318040 \times 10^{- 4} & 1.266611 \times 10^{- 3} & 9.907891 \times 10^{- 1} \end{matrix}] \end{matrix}$

Table 4. Matching results of three sets of images
View table
Table 4. Matching results of three sets of images
Scene Matching points of inertial navigationaided image feature matching method Matching pointsof RANSAC Accuracy rate /%
Experimental table 145 134 92.4
Laboratory 293 283 96.6
Office 171 161 94.2