文献类型：会议论文

英文题名：A New Traffic Signs Classification Approach Based on Local and Global Features Extraction

作者：He, Xiaojun[1];Dai, Benqi[1]

机构：[1]East China Univ Sci & Technol, Sch Informat & Control Engn, Shanghai, Peoples R China

会议论文集：6th International Conference on Information Communication and Management (ICICM)

会议日期：OCT 29-31, 2016

会议地点：Univ Hertfordshire, Hatfield, ENGLAND

主办单位：Univ Hertfordshire

语种：英文

外文关键词：traffic signs classification; LBP; CSLBP; DWT

摘要：In recent years, the automatic traffic signs recognition(TSR) has attracted researches' attention and it becomes a great challenge in Intelligent Transport System (ITS). Because of complex environmental or weather reasons, it is difficult for TSR to achieve high correct recognition rate and meet real-time at the same time. This paper focus on the traffic signs feature extraction and proposes a new effective approach to recognize the traffic signs with combining local feature and global feature. Center Symmetry Local Binary Pattern (CSLBP) is the improvement of Local Binary Pattern(LBP), We modify the CSLBP with multi sampling as the local feature. The global feature is the low frequency coefficients of Discrete Wavelet Transform (DWT) which owns good multi resolution ability. After extracted the two kinds of feature, we cascade them as the new feature to represent the traffic signs image. The final feature as the input data to classify the traffic signs with Support Vector Machine (SVM). In the GTSRB database, the result shows that the proposed approach achieves a accuracy of 97.67% which is superior to single feature approach and greatly reduces the recognition time.