Topics in matrix analysis
Data Fusion for Sensory Information Processing Systems
Data Fusion for Sensory Information Processing Systems
Towards Real-Time Cue Integration by Using Partial Results
ECCV '02 Proceedings of the 7th European Conference on Computer Vision-Part IV
COLD: The CoSy Localization Database
International Journal of Robotics Research
Sparse kernel SVMs via cutting-plane training
Machine Learning
Semantic place classification of indoor environments with mobile robots using boosting
AAAI'05 Proceedings of the 20th national conference on Artificial intelligence - Volume 3
Multi-modal Semantic Place Classification
International Journal of Robotics Research
Cue integration through discriminative accumulation
CVPR'04 Proceedings of the 2004 IEEE computer society conference on Computer vision and pattern recognition
LIBSVM: A library for support vector machines
ACM Transactions on Intelligent Systems and Technology (TIST)
An extended-HCT semantic description for visual place recognition
International Journal of Robotics Research
Image Segmentation by Probabilistic Bottom-Up Aggregation and Cue Integration
IEEE Transactions on Pattern Analysis and Machine Intelligence
SURF: speeded up robust features
ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part I
Sampling strategies for bag-of-features image classification
ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part IV
Bubble space and place representation in topological maps
International Journal of Robotics Research
| Hi-index | 0.00 |
This paper presents a new approach to the integration of different sensory cues in bubble space for place recognition. In bubble space, bubble surfaces enable the representation of all features in a manner that is implicitly dependent on robot pose while preserving their local S2-geometry. In the proposed approach, for each place, distinct groups of bubble surfaces conduce different cue descriptors which are then combined together. Unlike most previous work, merging cues of different nature is very simple regardless of the number of observations associated with each cue. Comparative experiments on a benchmark dataset indicate that while learning times are decreased considerably, recognition rates are comparable to state-of-the art approaches in place recognition.