Cascaded Segmentation Detection Networks for Text Based Traffic Sign Detection

Abstract:

           Since texts in traffic signs have precise and rich semantic information related to traffic condition, text-based traffic sign detection plays an important role in the autonomous driving and auxiliary driving of Intelligent Transportation System, just as traffic congestion detection, road detection, etc. The main difference between scene text detection and text-based traffic sign detection is that texts often fall on the traffic signs, so that the backgrounds of the texts in traffic sign are relatively clean. A new cascaded segmentation detection framework which is used to trained for text-based traffic sign detection. In this proposed system is to accurately detect the texts in traffic signs with high efficiency, fully avoiding the influence of background texts and symbol-based traffic signs. First traffic signs are segmented by using ROI (Region of Interest) to get the coarse area of traffic signs. Then the deep belief network is used in order to classify the text based traffic sign.

Proposed System;

               In this proposed system is to accurately detect the texts in traffic signs with high efficiency, fully avoiding the influence of background texts and symbol-based traffic signs. First, traffic signs are segmented by using ROI (Region of Interest)) to get the coarse area of traffic signs. Then the deep belief network is used in order to classify the text based traffic sign.