Pose Estimation Using April Tags for the Picking Problem

Abstract

This paper introduces a new method of pose estimation for the picking and stowing problems. The approach uses April tags to estimate an item’s pose from a single image. In testing, the approach has a root mean square error of 4.5 mm against a SIFT and RANSAC method on non-deformable items used in the 2015 Amazon Picking Challenge. This method could be valuable for stocking warehouse shelves and for other applications. 2