this is for holding javascript data
Volker Strobel edited subsection_Optical_Flow_label_sec__.tex
over 7 years ago
Commit id: ce6f5106ee12442440ce1e06ca0ebd111492b30f
deletions | additions
diff --git a/subsection_Optical_Flow_label_sec__.tex b/subsection_Optical_Flow_label_sec__.tex
index 40d2943..2cd8c1e 100644
--- a/subsection_Optical_Flow_label_sec__.tex
+++ b/subsection_Optical_Flow_label_sec__.tex
...
\subsection{Optical Flow}
\label{sec:opticalflow}
Optical flow algorithms are biologically inspired methods for navigation---taking inspiration from insects and birds~\cite{ruffier2003bio}.
They estimate the motion based on the shift of corresponding image keypoints in successive images. Gradient based approaches, such as the Lucas-Kanade method, keypoint-based methods, and more specific methods have been put forth.
Optical flow estimates the motion based on successive images: the shift of corresponding image keypoints in $x,y$-direction. The
approach is approaches are computationally rather complex.
It is a relative They belong to the class of local localization
method: without an initial reference point, it techniques and can only estimate the
distance position relative to an initial reference
point but not its absolute position in space. point. The
approach suffers approaches suffer from
drift: since the distance travelled is based on comparing successive images, errors are accumulating
errors over
time. The approach alone does time and typically do not
have the the possibility to correct provide a means for correcting these errors.
\citet{chao2013survey} compare different optical flow algorithms for the use with UAV navigation.
To render on-board optical flow estimation feasible for small MAVs, \citet{mcguire2016local} introduce a lightweight
optical flow algorithm that is able to run on-board of an MAV. algorithm. The algorithm uses compressed representations of images in the form of edge histogram to calculate the flow.