A Method of US Traffic Sign Detection and Recognition

ورقة مؤتمر

2012

Safran, Mejdl . 2012

A key issue in designing both autonomous vehicles and driver support systems is how to detect and recognize the traffic rules. One of these rules is the traffic signs. In this paper, we proposed a method for detecting and recognizing the United States regulatory traffic signs (stop sign, no-left-turn sign, no-right-turn sign, do-not-enter sign, and yield sign) based on color segmentation and shape analysis in real street-view images. Images that are taken as inputs by the proposed method are processed through three main phases: color segmentation, shape classification, and recognition of traffic signs. For this study, HSV color model is adopted for color segmentation which gives more accurate results even with high-lights and illumination changes. Shape signatures are used for shape classification. A ration matching technique using a decision tree is applied for the recognition phase. The proposed method is tested on real street-view images taken from Google Maps. Experimental results show that our approach is valid and can be applied for realtime applications.

• مدينة النشر: Carbondale
• موقع المؤتمر: Carbondale, IL
• المدرسة: Southern Illinois University Carbondale