基于高稳定SURF特征的交通标志识别

Traffic sign recognition based on stable SURFs

针对光照变化和部分遮挡这两种情形,提出一种基于多帧视频图像的高稳定特征的交通标志识别方法。利用有交通标志的多帧视频图像的SURF特征建立bag of SURFs特征向量集,与标准交通标志图像的模板特征向量集匹配,采用权值计分策略的最高得分确定交通标志的识别结果。对三种情形下的公开视频图像集进行了实验并与最新方法进行对比分析,结果表明新方法的交通标志识别效果具有明显的优越性,是在光照变化和部分遮挡情形下一种有效的交通标志识别方法。

This paper proposed a new method to recognize traffic signs based on SURFs with high stability. A set of SURFs, which was called bag of SURFs,was constructed using SURFs from multiple video frames that contain a traffic sign, and then matched with a set of SURFs from a standard traffic sign image. The recognition result was determined by the highest score of a weighted scoring scheme. Carried out a experiments based on a public dataset under three given conditions, and also performed comparative analysis with two recent methods. The results demonstrate that the method is an efficent traffic sign recognition method and yields a significant superiority under variable lighting or partial condition.