A recursive least squares solution for recovering robust planar homographies
TAROS'11 Proceedings of the 12th Annual conference on Towards autonomous robotic systems
Hi-index | 0.00 |
In this paper, a technique to design a robust feature extractor and descriptor for visual map building is proposed. The extracted features are required to be computationally attractive and invariant to image rotation, scale change and illumination. We adapted the Scale Invariant Features Transform (SIFT) algorithm for Map Building applications. Our main contributions are: firstly, we introduce of an adaptive version of the SIFT algorithm suitable for different visual perceptual environments. Secondly, we use of the L-infinity norm as a criterion for feature matching, which ensures more robustness against noises and uncertainties. Finally, we propose a new criterion to select the most stable features in order to improve the visual map building performances. Results based on real images shows the good performance obtained with the proposed approach.