Visual Detection, Tracking and Pose Estimation of a Robotic Arm End Effector

Abstract

In this work we present a robust tracking and pose estimation of an end effector of a 7 DoF robotic arm under a binocular stand-alone configuration. The proposed method is composed of sequence of three processes. The first one is a fast segmentation in HSV color space of a planar patch placed at the end effector. This step considers the segmentation using pixels classification and Euclidean distance a similarity measure. The second step is the selection of a region of interest, feature point extraction and tracking over that region. Finally, the feature points can be used to estimate the homography between world reference frame and image frame. The proposed methodology runs in real time and it is applicable for visual servoing and the grasping.