Unscented FastSLAM for UAV

Obstacle Detection Using Monocular Camera for Low Flying Unmanned Aerial Vehicle

“…• Landmark position errors on the Y axis are also proportional to their ground truth positions on X with the slope and polarity dependent on the amount and polarity of the error in c x (Figure 6.13, top plot, subplot [2,1]).…”

“…Therefore, the simulation evaluated the effect of lens distortion varying from 0% to 120% of the base value to estimate the amount of error resulting from ignoring the lens distortion. Figure 6.20, subplots [1,2] and [1,3]). The further the landmarks lay from the X axis, the bigger the errors they suffer.…”

“…Landmark positions on the Y and Z axes have less mean error, but the standard deviation are bigger. Figure 6.20 subplots [2,2] and [3,3] show that axes. This is exactly the kind of error that radial distortion corrects for.…”

“…In recent years, optical sensors are actively being incorporated into SLAM algorithms and successfully used in ground vehicle navigation. For aerial vehicles, the experiments are mostly limited to simulation [1] [2] [3] [4], and results from realistic aerial video data are rare.…”