Symmetry-based recognition of vehicle rears
Pattern Recognition Letters
Artificial neural networks in real-time car detection and tracking applications
Pattern Recognition Letters - Special issue on neural networks for computer vision applications
Saliency, Scale and Image Description
International Journal of Computer Vision
A Tutorial on Support Vector Machines for Pattern Recognition
Data Mining and Knowledge Discovery
Combining Local and Global Image Features for Object Class Recognition
CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Workshops - Volume 03
Journal of Cognitive Neuroscience
A multi-class object classifier using boosted Gaussian mixture model
ICONIP'10 Proceedings of the 17th international conference on Neural information processing: theory and algorithms - Volume Part I
| Hi-index | 0.00 |
We propose a novel obstacle categorization model combining global feature with local feature to identify cars, pedestrians and unknown backgrounds. A new obstacle identification method, which is hybrid the global feature and local feature, is proposed for robustly recognizing an obstacle with and without occlusion. For the global analysis, we propose the modified GIST based on biologically motivated the C1 feature, which is robust to image translation. We also propose the local feature based categorization model for recognizing partially occluded obstacle. The local feature is composed of orientation information at a salient position based on the C1 feature. A classifier based on the Support Vector Machine (SVM) is designed to classify these two features as cars, pedestrians and unknown backgrounds. Finally, all classified results are combined. Mainly, the obstacle categorization model makes a decision based on the global feature analysis. Since the global feature cannot express partially occluded obstacle, the local feature based model verifies the result of the global feature based model when the result is an unknown background. Experimental results show that the proposed model successfully categorizes obstacles including partially occluded obstacles.