01137496

Component

Road sign plays an important role in highway management by providing the drivers and road users guidance, warning and other driving related information. Assets of road signs represent a substantial investment of roadway agencies. Proper sign maintenance and inventory is an essential tool for infrastructure management and maintenance. Currently, sign inventory is mostly conducted by human observation of the digital images of roadway scene, or commonly called Right-of-Way imaging. Automated system would substantially improve the processing speed and accuracy of two key processing tasks, sign detection and sign classification, since the human observation of the large amount of images is tedious, error-prone and time-consuming. This paper emphasizes on the study of sign classification. Classification is to categorize signs into proper classes, which is an important process and substantially more difficult for automation than sign detection in a sign inventory system. The most frequently used technique used in previous research for classification is neural network. However, neural network might suffer from the local minimum problem if the training data are not properly picked. In addition, neural network is lack of explainable inner theoretical rule which bring difficulty to fine tune the performance of the model. This paper presents a method which combines feature extraction, Support Vector Machine (SVM), and multi-class classification. SVM is a statistical learning method based on Vapnik-Chervonenkis (VC) dimension and structural minimum principle (1). It is designed to overcome the aforementioned two drawbacks of neural network method. The feature extraction was accomplished by Principal Component Analysis (PCA) which reduces the dimension of the image as well as keeping the most important features. Experimental result presented in this paper demonstrates that the SVM method has potential to solve the road sign classification problem.

01120148

09-3125

English

20p

2009

Transportation Research Board 88th Annual Meeting

DVD

Figures (4) ; References (25) ; Tables (1)

Algorithms; Asset management; Classification; Detection and identification systems; Image processing; Traffic control devices; Traffic signs

Data and Information Technology; Highways; I10: Economics and Administration

Transportation Research Board Annual Meeting 2009 Paper #09-3125

BTRIS, TRIS, TRB

Jan 30 2009 7:31PM