摘要
为了提高交通标志识别效率及准确性,在研究了深度神经网络、特征提取模基础上,设计了一种包含主干网络和辅助网络的孪生网络。孪生网络中主干网络和辅助网络使用相同的训练集。首先对主干网络进行训练,使其收敛于交通标志训练集;其次,通过知识提取帮助辅助网络进行训练,从而获得更好的交通标志识别率;最后,对辅助网络模型进行剪枝,从而降低总体网络计算成本。仿真分析结果表明,所提方法在保证一定识别率情况下,能够大幅降低网络参数。仿真结果进一步验证了所提系统的有效性。
In order to improve the efficiency and accuracy of traffic sign recognition,based on the research of deep neural network and feature extraction module,a twin network including main network and auxiliary network is designed.In the twin network,the backbone network and the auxiliary network use the same training set.Firstly,the backbone network is trained to converge to the traffic sign training set.Secondly,knowledge extraction is used to assist network training,so as to obtain better traffic sign recognition rate.Finally,the auxiliary network model is pruned to reduce the overall network computing cost.Through the simulation analysis,the results show that the proposed method can greatly reduce the network parameters while ensuring a certain recognition rate.The simulation results further verify the effectiveness of the proposed system.
作者
刘丽景
Liu Lijing(Xi’an Peihua University,Xi’an 710025,China)
出处
《单片机与嵌入式系统应用》
2021年第11期14-17,共4页
Microcontrollers & Embedded Systems
基金
西安培华学院2020年度校级科研项目——基于实例分割的街景图像自动标注策略的研究(PHKT2007)。
关键词
交通标志
深度学习
孪生网络
知识提取
traffic signs
deep learning
twin networks
knowledge extraction
