Applications to Safe Human–Robot Interaction - 3D Computer Vision: Efficient Methods and Applications - page 276

Chapter 7
Applications to Safe Human-Robot Interaction
This chapter addresses the scenario of safe human-robot interaction in the industrial
production environment. For example, in car manufacturing, industrial production
processes are characterised by either fully automatic production sequences carried
out solely by industrial robots or fully manual assembly steps where only humans
work together on the same task. Close collaboration between humans and industrial
robots is very limited and usually not possible due to safety concerns. Industrial
production processes may increase their efficiency by establishing a close collabo-
ration of humans and machines exploiting their unique capabilities, which requires
sophisticated techniques for human-robot interaction. In this context, the recogni-
tion of interactions between humans and industrial robots requires vision methods
for three-dimensional pose estimation and tracking of the motion of human body
parts based on three-dimensional scene analysis.
Section 7.1 gives an overview of vision-based safe human-robot interaction and
illustrates the importance of gesture recognition in the general context of human-
robot interaction. In Sect. 7.2 the performance of the three-dimensional approach
of Schmidt et al. ( 2007 ) in the detection and tracking of objects in point clouds (cf.
Sect. 2.3 ) in a typical industrial production environment is evaluated. In Sects. 7.3 -
7.5 , the methods of Barrois and Wöhler ( 2008 ) and Hahn et al. ( 2010a , 2010b )for
the three-dimensional detection, pose estimation, and tracking of human body parts
(cf. Sect. 2.2.1.2 ) as well as the approach suggested by Hahn et al. ( 2009 , 2010b )
for recognising the performed actions are evaluated in similar scenarios.
7.1 Vision-Based Human-Robot Interaction
Vision-based systems for detecting the presence of humans in the workspace of an
industrial robot and safeguarding the cooperation between a human and an industrial
robot are described in Sect. 7.1.1 . These systems take into account the approximate
shape of the human but do not determine the motion behaviour of human body
parts. Section 7.1.2 provides an overview of methods for the recognition of ges-