This paper proposes a scant overlap registration strategy for 3D (three-dimensional) reconstruction of
indoor map by using a robust, fast and efficient approach. First, a design methodology based on the use of an
omnidirectional robot is proposed for the large-deformation data and scant overlap. Second, this robot uses
vision-servo approach by a Kinect device for the purposes of avoiding obstacles and providing 3D data for
walls. Third, experimental results illustrate that map’s data sets with scant overlap cannot be registered directly
by traditional registration methods. Finally, this study is based on the bee-colony algorithm to solve this
registration problem.
