Unsupervised Image Segmentation Using A Simple MRF Model with A New Implementation Scheme
ICPR '04 Proceedings of the Pattern Recognition, 17th International Conference on (ICPR'04) Volume 2 - Volume 02
IROS'09 Proceedings of the 2009 IEEE/RSJ international conference on Intelligent robots and systems
ETFA'09 Proceedings of the 14th IEEE international conference on Emerging technologies & factory automation
| Hi-index | 0.00 |
This paper presents an approach of an extended terrain classification procedure for an autonomous mobile robot with multimodal features. Terrain classification is an important task in the field of outdoor robotics as it is essential for negotiability analysis and path planning. In this paper I present a novel approach of combining multi modal features and a Markov random field to solve the terrain classification problem. The presented model uses features extracted from 3D laser range measurements and images and is adapted from a Markov random field used for image segmentation. Three different labels can be assigned to the terrain describing the classes road, for easy to pass flat ground, rough for hard to pass ground like grass or a field and obstacle for terrain which needs to be avoided. Experiments showed that the algorithm is fast enough for real time applications and that the classes road and street are detected with a rate of about 90% in rural environments.