基于多特征组合的交通标识识别

J4 ›› 2015, Vol. 37 ›› Issue (04): 776-782.

基于多特征组合的交通标识识别

齐朗晔1,2,3,张重阳1,2,何成东3

(1.南京理工大学计算机科学与工程学院，江苏南京 210094；
2.南京理工大学高维信息智能感知与系统教育部重点实验室，江苏南京 210094；
3.中国电子科技集团公司第五十二研究所物联网研究院，浙江杭州 310012)

收稿日期:2013-08-28 修回日期:2014-04-03 出版日期:2015-04-25 发布日期:2015-04-25
基金资助:
国家自然科学基金资助项目(90820306)；高维信息智能感知与系统教育部重点实验室（南京理工大学）开放基金资助项目(MEKL201308)

Traffic sign recognition based on multifeature combination

QI Langye1,2,3,ZHANG Chongyang1,2,HE Chengdong3

(1.School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing 210094;
2.Key Laboratory of Intelligent Perception and Systems for HighDimensional Information of Ministry of Education,
Nanjing University of Science and Technology,Nanjing 210094;
3.Institute of IoT, No 52 Research Institute,China Electronics Technology Group Corporation,Hangzhou 310012,China)

Received:2013-08-28 Revised:2014-04-03 Online:2015-04-25 Published:2015-04-25

摘要/Abstract

摘要：

在分块核函数的基础上提出了基于多个图像特征进行组合决策的识别方法。该算法先对交通标识图像提取两个不同的特征，即HOG特征和基于子模式组合的分块核函数特征，然后针对不同特征构造相应的分类器，最后对这几个分类器的输出采用投票机制进行决策融合。在德国交通标识数据库上的实验结果表明，该方法相比单特征识别具有更高的识别准确率。

关键词: 核Fisher非线性鉴别分析, 特征组合, 分块核方法, 交通标识识别

Abstract:

We present a recognition method based on multiple graphic features to do decision combination on the basis of block kernel function.Firstly,the algorithm extracts two different characteristics of traffic signs, i.e. HOG features and the feature of block kernel function based on submode combination.Secondly,we construct different classifiers for different features accordingly. Finally, the decision fusion is done by the voting mechanism for the outputs of both classifiers.Experiments on Germany Traffic Signs Database prove that our proposal has a higher recognition accuracy compared with the single-feature recognition.

Key words: kernel Fisher nonlinear discriminate analysis;feature combination;block kernel function;traffic sign recognition

齐朗晔1,2,3,张重阳1,2,何成东3. 基于多特征组合的交通标识识别[J]. J4, 2015, 37(04): 776-782.

QI Langye1,2,3,ZHANG Chongyang1,2,HE Chengdong3. Traffic sign recognition based on multifeature combination [J]. J4, 2015, 37(04): 776-782.