Tracking traffic signs in video sequences based on a vehicle velocity

The paper proposes an efficient algorithm for road sign detection in a video obtained from a car dash camera. The algorithms for road sign detection and recognition are implemented using CUDA and operate in real time. A vehicle velocity is used to predict the road sign position in adjacent frames in a video sequence. The paper shows that the road sign tracking of this type improves the system reliability. The experimental results have confirmed high efficiency of the developed road sign detection system.