Low-cost and Faster Tracking Systems Using Core-sets for Pose-Estimation.
Year: 2015
Type: preprint
Abstract: How can a \$20 toy quadcopter hover using a weak Internet of Things mini-board and a web-cam? In the pose-estimation problem we need to rotate a set of $n$ marker (points) and choose one of their n! permutations, so that the sum of squared corresponding distances to another ordered set of $n$ markers is minimized. A popular heuristic for this problem is ICP.
We prove that \emph{every} set has a weighted subset (core-set) of constant size (independent of $n$), such that computing the optimal orientation of the small core-set would yield \emph{exactly} the same result as using the full set of $n$ markers. A deterministic algorithm for computing this core-set in $O(n)$ time is provided, using the Caratheodory Theorem from computational geometry.
We developed a system that enables low-cost and real-time tracking by computing this core-set on the cloud in one thread, and uses the last computed core-set locally in a parallel thread. Our experimental results show how these core-sets can boost the tracking time and quality for large and even small sets of both IR and RGB markers on a toy quadcopter. Open source code for the system and algorithm is provided
Source: arXiv (Cornell University)
Authors: Soliman Nasser, Ibrahim Jubran, Dan Feldman
Language: English
Topic: Advanced Vision and Imaging
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2277927960