| Summary: | To develop systems for letting the robot receive human orders precisely in the daily work space, this paper proposes an experimental system for finger direction recognition (FDR) in 3D space using stereo matching by two cameras mounted on the robot. In this FDR method, our algorithm estimates two finger directions: one of them is finger direction theta on the photographic surface; the other is finger direction Phi of depth. These directions are obtained as projections of finger direction in 3D space to two planes. This FDR method consists of three steps: (1) detecting hand contour, (2) estimating theta, and (3) estimating Phi. To evaluate this system for FDR in 3D space, we perform a series of evaluation experiments using a turntable capable of fixing a hand on a specific finger direction. We tried to estimate various finger directions and various distances for two major pointing hand forms, and evaluated the precision of theta and Phi. As a result, we conclude that theta and Phi estimations are valid when the distance between the camera and object is less than 140 and 110 cm for theta and Phi, respectively. (C) 2012 Masahiro Ohka. Published by Elsevier Ltd. Selection and/or peer-review under responsibility of the Centre of Humanoid Robots and Bio-Sensor (HuRoBs), Faculty of Mechanical Engineering, Universiti Teknologi MARA.
|
|---|
