Authors:
David Forslund
;
Per Cronvall
and
Jacob Roll
Affiliation:
Autoliv Electronics AB, Sweden
Keyword(s):
Scene classification, Bag of Words, Visual Words.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Biomedical Engineering
;
Biomedical Signal Processing
;
Computer Vision, Visualization and Computer Graphics
;
Data Manipulation
;
Feature Extraction
;
Features Extraction
;
Health Engineering and Technology Applications
;
Human-Computer Interaction
;
Image and Video Analysis
;
Informatics in Control, Automation and Robotics
;
Methodologies and Methods
;
Motion, Tracking and Stereo Vision
;
Neurocomputing
;
Neurotechnology, Electronics and Informatics
;
Pattern Recognition
;
Physiological Computing Systems
;
Real-Time Vision
;
Sensor Networks
;
Signal Processing, Sensors, Systems Modeling and Control
;
Soft Computing
Abstract:
This paper aims at employing scene classification in real-time to the two-class problem of separating city and rural scenes in images constructed from an infrared sensor that is mounted at the front of a vehicle. The 'Bag of Words' algorithm for image representation has been evaluated and compared to two low-level methods 'Edge Direction Histograms', and 'Invariant Moments'. A method for fast scene classification using the Bag of Words algorithm is proposed using a grey patch based algorithm for image element representation and a modified floating search for visual word selection. It is also shown empirically that floating search for visual word selection outperforms the currently popular k-means clustering for small vocabulary sizes.