Pose Estimation for Mobile Robots Working on Turbine Blade

X. Ma, Q. Chen, J. Liu, Z. Sun, and W. Zhang (PRC)


Mobile robots, pose estimation, feature detection, SIFT, RANSAC


For mobile robots working on giant turbine blades, pose estimation of the robots is very important in their tasks. Vision based scheme utilizes only the visual information of the robot’s surrounding environment, thus becoming a good candidate scheme for pose estimation. A most important task in this scheme is to detect feature points in sucessive image frames and to match them. In this paper, an improve pose estimation algorithm based on SIFT (Scale Invariant Feature Transform) is presented considering the characteristics of local images of the turbine blades, and the pose estimation problem and condition. The improvement includes pre-subsampling the image, which reduces the computation, and bidirectional matching, which improves the precision. RANSAC (Random Sample Consensus) method is used to get a better estimation of the robot’s pose. The experiment platform is built and the experimental results show the validity of the proposed algorithm.

Important Links:

Go Back