摘要
传统的基于卷积神经网络的车型识别算法存在识别相似车型的准确率不高,以及在网络训练时只能使用图像的灰度图从而丢失了图像的颜色信息等缺陷。对此,提出一种基于深度卷积神经网络(Deep Convolution Neural Network,DCNN)的提取图像特征的方法,运用深度卷积神经网络对背景较复杂的车型进行网络训练,以达到识别车型的目的。文中采用先进的深度学习框架Caffe,基于AlexNet结构提出了深度卷积神经网络的模型,分别对车型的图像进行训练,并与传统CNN算法进行比较。实验结果显示,DCNN网络模型的准确率达到了96.9%,比其他算法的准确率更高。
The accuracy of traditional convolution neural network recognizing the vehicle model is not high when recognizing similar models,and the gray scale of image can only be used in the network training with the loss of color information of the image.Based on this,a method of extracting image features based on deep convolution neural network（DCNN）was proposed.The deep convolution neural network is used to carry out network training for the vehicle model with complex background,so as to achieve the purpose of recognizing models.In this paper,by using advanced deep learning framework Caffe,a deep convolution neural network based on AlexNet structure was proposed with the training of the image of vehicle model and the comparison with traditional convolution neural network.The experimental results show that the accuracy rate of DCNN network model can reach 96.9% with a higher accuracy.
作者
石磊
王亚敏
曹仰杰
卫琳
SHI Lei;WANG Ya-min;CAO Yang-jie;WEI Lin(School of Information Engineering,Zhengzhou University, Zhengzhou 450001,China;School of Software and Applied Science and Technology,Zhengzhou University, Zhengzhou 450002, China)
出处
《计算机科学》
CSCD
北大核心
2018年第5期280-284,共5页
Computer Science
基金
国家自然科学基金(U1304603)
河南省高等学校重点科研项目(17A520016)
郑州大学优秀青年教师发展基金(1521337044)资助
关键词
卷积神经网络
车型识别
深度学习
Convolution neural network
Vehicle identification
Deep learning
