摘要
道路交通标志识别作为主动安全驾驶系统和自动驾驶系统的重要组成部分,在道路行车安全过程中作用巨大,而真实场景采集的交通标志易受复杂环境的影响,造成自动识别精度不高。针对这一问题,以LeNet-5模型为基础,结合密集连接卷积神经网络(Dense Convolutional Network,DenseNet)思想进行网络结构调整,通过图像特征重用挖掘网络潜力、防止梯度消失,加入Dropout策略改善过拟合现象,利用正则化机制约束模型特性,提高模型的泛化能力。在德国交通标志数据集(German traffic sign recognition benchmark,GTSRB)上进行实验,构建了基于特征重用的交通标志识别系统。
As an important part of active safe driving system and self-driving system,traffic sign recognition plays an important role in the process of road driving safety,however traffic signs collected in real scenes are susceptible to complex environment,resulting in low recognition accuracy.In response to this problem,the network structure is adjusted based on lenet-5 model and densenet ideas,mining network potential and preventing vanishing gradient through feature reuse,the Dropout strategy was added to improve the over-fitting phenomenon,the regularization mechanism was used to constrain the model characteristics and improve generalization ability of the model.Experiments were conducted on the German traffic sign recognition benchmark(gtsrb),a traffic sign recognition system based on feature reuse is constructed.
作者
仲会娟
ZHONG Hui-juan(College Ofartificial Intelligence,Yango University,Fuzhou Fujian 350015)
出处
《数字技术与应用》
2020年第2期24-26,共3页
Digital Technology & Application
基金
2018年福建省中青年教师教育科研项目(JT180724)。
