欢迎访问《图学学报》 分享到:

图学学报

• 专论:第16届媒体智能与大数据计算会议(CIDE & DEA 2019 大连) • 上一篇    下一篇

交通标志识别特征提取研究综述

  

  1. (大连民族大学计算机科学与工程学院,辽宁 大连 116605)
  • 出版日期:2019-12-31 发布日期:2020-01-20
  • 基金资助:
    国家自然科学基金项目(6130089);辽宁省自然基金项目(201602199);辽宁省高等学校创新人才支持计划(LR2016071);国家级大学生创 新训练项目(201912026022)

Review on Feature Extraction of Traffic Sign Recognition

  1. (College of Computer Science and Engineering, Dalian Minzu University, Dalian Liaoning 116605, China)
  • Online:2019-12-31 Published:2020-01-20

摘要: 交通标志识别(TSR)是智能交通系统(ITS)的一个重要研究方向,而特征提取是交 通标志识别研究中的重点。聚焦交通标志识别的特征提取,综述了常见的人工特征(颜色直方图、 尺度不变特征变换特征、局部二值模式特征、方向梯度直方图特征、Haar-like 特征、Gabor 小 波特征、Canny 特征等)和深度特征(提取自 AlexNet,VGG16,Inception 等),并在同一数据集 (GTSRB)上提取多种特征,采用相同分类器,通过相同评价指标体系进行定量比较与分析,并 以图表方式,针对不同特征和不同交通标志类别,进行直观的性能比较研究,以期为交通标志 识别时特征向量的选择和深入研究提供参考。

关键词: TSR, ITS, 特征提取, 人工特征, 深度特征

Abstract: Traffic sign recognition (TSR) is an important research direction of intelligent transportation system (ITS). Feature extraction is the key point of traffic sign recognition research. This paper focuses on feature extraction of traffic sign recognition and summarizes common manual features and depth features. Manual features include color histogram, scale invariant feature transformation feature, local binary pattern feature, directional gradient histogram feature, Haar-like feature, Gabor wavelet feature, Canny feature, etc. Depth features are extracted from AlexNet, VGG16, Inception, etc. Various features are extracted from the same data set (GTSRB). Various features are compared and analyzed quantitatively by using the same classifier and the same evaluation index system. This paper makes an intuitive comparative research of performance for different features and different types of traffic signs by means of charts and graphs, aiming at providing a reference for the selection of feature vectors and for the further research of traffic sign recognition.

Key words: TSR, ITS, feature extraction, manual features, depth features