BROWSE
Detail
Conference Paper
Automatic Subway Sign Detection and Recognition Based on SVMs
Cited - time in 
Share 
Share
- Authors
- Ho-Sub Yoon, DongJin Lee, Jae-Hong Kim, Myung-Ae Chung, Sul-Ho Kim
- Issue Date
- 2013-06
- Citation
- International Conference on Computer Applications and Information Processing Technology (CAIPT) 2013, pp.1-4
- Language
- English
- Type
- Conference Paper
- Abstract
- In this paper, we present a walking guidance system based on Support Vector Machines (SVM) with HOG features for the visually impaired in subway stations. This system automatically detects and recognizes several signs such as restroom, elevator, ticket office, and etc. from the natural scenes, hence visuallyhandicapped people could utilize the system to find their own way (e.g., using information of ticket office and direction) in the subway stations. The proposed walking guidance system mainly consists of three stages. 1) sign detection by the rule based modeling using a prior knowledge; 2) feature extraction method using HOG(Histogram Oriented Gradients) vectors; 3) SVM classifier to discriminate between signs and nonsigns; Experiment results show a reasonable success rate and a false positive rate (FAR) of 0.0001% to distinguish a sign from a non-sign (noise).
- KSP Keywords
- Detection and Recognition, Experiment results, False Positive(FP), False Positive Rate, Guidance System, Histogram Oriented Gradients, HoG features, Rule-Based Modeling, SVM Classifier, Sign detection, Success rate