摘要
本文提出一种基于卷积神经网络的交通灯识别方法,解决传统交通灯识别算法在复杂交通场景中存在稳定性差、准确率低等问题.本方法采用深层神经网络模型作为分类器,将数据集中的彩色图像经预处理后作为卷积神经网络的输入,自动提取特征,最后通过多模式预测算法得出识别结果.通过多组实验,证明了本文方法的有效性及实用性.
Traditional method of traffic light recognition suffers from poor robustness and low accuracy in complex traffic scenes. To solve these problems, this paper proposes a method of traffic light recognition based on convolutional neural network. We use deep convolutional neural network as the classifier. The in- put of this network is a series of pre-processed color images. The network can extract features automatically, and the final result is produced by multi-pattern prediction method. The experimental results show that the proposed method is effective and practical.
作者
贾瑞明
刘立强
JIA Ruiming;LIU Liqiang(Col.of Electronic Information Engineering,North China Univ.of Teeh.,100144,Beijing,China)
出处
《北方工业大学学报》
2018年第5期22-30,共9页
Journal of North China University of Technology
基金
北方工业大学学生科技活动项目
关键词
交通灯识别
卷积神经网络
多模式预测
traffic light recognition
convolutional neural network
multi-pattern prediction
