Visual localization and object tracking for the NAO robot in dynamic environment

Abstract

In this paper, we present an integrated approach of multiple algorithms for visual localization and object tracking. First, a spatial point monocular vision range model is established. Combined with vision technology, we can deduce a precise position of the target in the world frame, and the region of recognized object is regarded as the real tracking region. Second, The Camshift/Kalman/Particle algorithm has been fused to resolve some common problems in video tracking, such as background interference, target with sudden move, occlusion, dim-small size and etc. Finally, a monocular vision range experiment and a target object tracking experiment are carried out. In order to complete the autonomous navigation experiment, both the visual localization method and target object tracking algorithm are implemented on the robot. These experimental results show the validity of our approaches.