摘要
提出了一种能够有效克服曝光不均复杂环境下的交通标志识别方法.采用改进的多尺度Log Gabor小波进行交通标志的多分辨率特征提取,根据不同尺度下的特征信息进行相位一致性计算,提取能够有效克服曝光影响的目标相位信息,通过优化的多分类支持向量机(SVM)进行多目标分类,并在德国交通标志标准数据库(GTSRB)上进行测试.结果表明:本方法对6大类主要交通标志样本的平均识别率达到98%,优于传统方法;在Intel双核CPU 2.4 GHz计算机平台上,本方法对数据库中不同尺度的图片处理速度达到28帧·s-1,满足了实时性要求;克服了光照不均的问题,适用于复杂条件下的交通标志识别,能够满足鲁棒性需求.
A traffic sign recognition method was proposed to solve the nonuniform exposure in complex traffic environments. The improved Log Gabor wavelet was used to extract feature of traffic signs at multi scales. The phase consistency was calculated to overcome the influence of environmental illumination. Support vector machine (SVM) was used as muhiple targets classifier to be tested in the authoritative traf fie sign database of German traffic sign recognition benchmark (GTSRB). The results show that the average recognition rate of the proposed method for 6 main category is 98%, which is better than those of the tradi- tional algorithms. The computational efficiency with Intel dual core CPU 2.4 GHz is 28 frames . s-1, which satisfies the requirement in real-time. The proposed method overcomes the problem of uneven ex- posure in complex traffic environment, which can achieve good robustness.
出处
《江苏大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2013年第5期514-518,共5页
Journal of Jiangsu University:Natural Science Edition
基金
国家自然科学基金资助项目(50675099)
江苏省普通高校研究生创新基金资助项目(cxlx11_0180)