Real-Time Detection and Tracking of Traffic Sign in Video Sequences for Autonomous Mobile Robot

In this paper, we discuss theoretical foundations and a practical realization of real-time traffic sign detection and tracking method for a autonomous vehicle (P3-AT mobile robot). Continuous adaptive mean shift (Cam-Shift) algorithm is efficient for object tracking with its high speed and insensitiveness to the rotation and scale of the target, but it is influenced by the background. In order to improve the tracking accuracy, we design a traffic sign probability model, and present a depth-first region search algorithm to remove the redundant background information, and then track the sign by Cam-Shift method. The experiments are conducted in a simulated urban environment with high fidelity. The experimental results show that the algorithm can improve the detection accuracy and tracking efficiency under the condition of traffic sign images with clutter background, or even partial occlusion. On the implementation side, with the core code implemented based on the OpenCV library, the results accomplish in real-time.