An Improved Anchor-Free Detection Method for Traffic Sign Detection

Abstract

As a basic task in the intelligent driving system, the traffic sign detection can locate and classify the traffic signs in real time and accurately. In response to the own design limitation of anchor-based detection methods, an improved anchor-free detection method is proposed. The method adds feature reinforcement module and head reinforcement module. First, in order to solve the small sign detection problem, a feature reinforcement module based on the hybrid attention mechanism is proposed. Secondly, in order to reduce the interference of complex background information to distinguish the prospects from the background, a head reinforcement module is proposed. We perform sufficient experiments on the CCTSDB dataset, and the experimental results demonstrate the effectiveness and robustness of the proposed method.