计算机应用 ›› 2020, Vol. 40 ›› Issue (10): 2872-2880.DOI: 10.11772/j.issn.1001-9081.2020020214

• 人工智能 • 上一篇    下一篇

基于改进空间金字塔池化卷积神经网络的交通标志识别

邓天民, 方芳, 周臻浩   

  1. 重庆交通大学 交通运输学院, 重庆 400074
  • 收稿日期:2020-03-02 修回日期:2020-05-21 出版日期:2020-10-10 发布日期:2020-05-27
  • 通讯作者: 方芳
  • 作者简介:邓天民(1979-),男,四川阆中人,副教授,博士,主要研究方向:交通大数据、机器人;方芳(1994-),女,四川自贡人,硕士研究生,主要研究方向:机器学习、交通环境感知;周臻浩(1995-),男,浙江宁波人,硕士研究生,主要研究方向:机器学习、交通环境感知。
  • 基金资助:
    国家自然科学基金资助项目(51678099);重庆市科技人才培养计划项目(CSTC2013KJRC-QNRC0148)。

Traffic sign recognition based on improved convolutional neural network with spatial pyramid pooling

DENG Tianmin, FANG Fang, ZHOU Zhenhao   

  1. College of Traffic and Transportation, Chongqing Jiaotong University, Chongqing 400074, China
  • Received:2020-03-02 Revised:2020-05-21 Online:2020-10-10 Published:2020-05-27
  • Supported by:
    This work is partially supported by the National Natural Science Foundation of China (51678099), the Science and Technology Talents Training Program of Chongqing (CSTC2013KJRC-QNRC0148).

摘要: 针对雾天、光照、遮挡和大倾角等因素导致的交通标志识别准确率低、泛化性差等问题,提出一种基于神经网络的轻量级交通标志识别方法。首先,利用图像归一化、仿射变换和限制对比度自适应直方图均衡化(CLAHE)方法进行图像预处理,以提高图像质量;其次,基于卷积神经网络(CNN),融合空间金字塔结构和批量归一化(BN)方法构建改进空间金字塔池化卷积神经网络(SPPN-CNN)模型,并利用Softmax分类器实现交通标志分类;最后,选用德国交通标志识别数据集(GTSRB),对比不同图像预处理方法、模型参数和模型结构的训练效果,并验证和测试所提模型。实验结果表明,SPPN-CNN模型的识别精度达到98.04%,损失小于0.1,在低配GPU条件下识别速率大于3 000 frame/s,验证了模型精度高、泛化性强、实时性好的特点。

关键词: 图像去雾, 空间金字塔池化, 卷积神经网络, Softmax分类器, 交通标志识别

Abstract: In order to solve the problems of low accuracy and poor generalization of traffic sign recognition caused by factors such as fog, light, occlusion and large inclination, a lightweight traffic sign recognition method based on neural network was proposed. First, in order to improve image quality, the methods of image normalization, affine transformation and Contrast Limited Adaptive Histogram Equalization (CLAHE) were used for image preprocessing. Second, based on Convolutional Neural Network (CNN), spatial pyramid structure and Batch Normalization (BN) were fused to construct an improved CNN with Spatial Pyramid Pooling (SPP) and BN (SPPN-CNN), and Softmax classifier was used to perform the traffic sign recognition. Finally, the German Traffic Sign Recognition Benchmark (GTSRB) was used to compare the training effects of different image preprocessing methods, model parameters and model structures, and to verify and test the proposed model. Experimental results show that for SPPN-CNN model, the recognition accuracy reaches 98.04% and the loss is less than 0.1, and the recognition rate is greater than 3 000 frame/s under the condition of GPU with low configuration,verifying that the SPPN-CNN model has high accuracy, strong generalization and good real-time performance.

Key words: image dehazing, Spatial Pyramid Pooling (SPP), Convolution Neural Network (CNN), Softmax classifier, traffic sign recognition

中图分类号: