一种多环境适用的交通标志识别模型

交通标志识别对于自动驾驶和辅助驾驶系统中非常重要,但大多数相关研究仅局限在白天场景的识别,若用于夜间,光强差异太大会导致识别准确率显著下降。为解决该问题,文中提出了光强分类模型,可根据光强强度划分场景识别标志牌,保证夜间较高的识别率。该模型通过KNN和SVM构造邻接矩阵和训练特征向量来判断分场景处理出的ROI,从而确定具体标志牌种类。实验证明,该模型在不同环境下识别准确率高达98.1%。

A Multi-environmental Traffic Sign Recognition Model

Traffic sign recognition plays an important role in autopilot and assisted driving systems. However, most related studies were limited to the identification of daylight scenarios. If these models were applied in night, the difference of light intensity would lead to a significant decrease in recognition accuracy. In order to solve this problem, this paper proposed a light intensity classification model, which used different methods to identify the signboard according to the difference of light intensity to ensure excellent recognition rates in night. Firstly, the model divided the scene and processed the ROI by different methods. And then, eigenvectors were trained by KNN and SVM. Finally, the trained eigenvectors could identify the type of a sign. Experiments proved that the accuracy rate of this model in different environments was as high as 98.1%.