Efficient Traffic Sign Detection Based on High Dynamic Range Preprocessing Module

Abstract

Traffic sign recognition systems are primordial for intelligent vehicles. Various vision based detection systems have been proposed to detect road panels, but few of them tackle adverse conditions of acquisition such as bad weather or high dynamic range scenes. In this paper, we introduce efficient traffic sign detection when facing difficult illumination conditions. First, a high dynamic range preprocessing module based on Neural Network exposure fusion is used. After that,  shape or color based detection phase is performed to detect regions of interest. Evaluation and comparisons using some relevant works showed that the added preprocessing module enhances the detection rate facing adverse illumination conditions.