Proposal of Artificial Mark to Measure 3D Pose by Monocular Vision

This paper proposes an artificial mark attached to objects for measuring 3D pose of the objects. We first proposes three marks with four point features. These differences are image processing method to extract the features : using edge extraction, using erosion, using color extraction. We compare error characteristic of these marks through measurement experiments. As results of the experiment, the method with color extraction, which has most simple image processing method, is most robust one in cases that a mark is measured from various pose (mean of error : 10.4 deg, mean of std. dev. : 7.06 deg, worst std. dev. : 24.7 deg) or is attached on surface. Moreover, the mark enables the shortest measurement time (0.478 s/6 marks). The usefulness of the mark with color extraction was verified through the success of the experiment that a manipulator handles three objects with the marks.