A Traffic Sign Detection Method with High Accuracy and Efficiency

摘要

Traffic sign recognition is a challenging computer vision problem for its high requirement of reliability and efficiency. In this paper, we propose a prohibitory traffic sign detection method. The method first extracts a red bitmap from the input image and then detects circles in the red bitmap. To detect circles more efficiently, a new circle detection method is proposed which is of high accuracy, low computational complexity and low memory consumption. Experimental results indicate that the proposed method can robustly detect prohibitory signs in all kinds of adverse situations including bad lighting condition, similar background color, multiple signs clustering, and partial occlusion. In addition, the method is efficient enough for real-time applications.