本文已被：浏览 1357次 下载 787次
投稿时间：2017-06-10
投稿时间：2017-06-10
中文摘要: 为了对车辆款式和型号进行分类筛选，降低侦查人员的劳动强度，提出了一种用约束卷积神经网络实现轿车款式识别的方法，相比车辆类型识别更为精细。首先通过测试进行正面"车脸"的识别，然后尝试整车车身（并带有部分背景）的识别。测试结果表明，对于包含部分背景内容的整车车身识别，在卷积神经网络中添加约束条件后，误识别率有4.06%的降幅，从而验证了该方法的有效性。
Abstract:A constrained convolutional neural network (CNN) is proposed to recognize and screen car models,which can reduce the labor strength of the investigators.Model recognition is in a grain finer than the recognition vehicle type (cars,trucks,buses,and so on).Front face recognition is firstly investigated,and then the whole body (with part of the background) recognition is tried.The test results show that,even for the vehicle body recognition with part of the background,the false recognition rates also have a decreasing amplitude of 4.06% under the configuration of constrained convolutional neural network.
keywords: convolutional neural network constraint variables vehicle appearance car model recognition whole car image
文章编号：20182017 中图分类号： 文献标志码：
基金项目:上海市科学技术委员会重点实验室专项基金（05DZ22305）。
| 作者 | 单位 | |
| 贾振堂 | 上海电力学院 电子与信息工程学院 | 462458081@qq.com |
| Author Name | Affiliation | |
| JIA Zhentang | School of Electronics and Information Engineering, Shanghai University of Electric Power, Shanghai 200090, China | 462458081@qq.com |
引用文本：
贾振堂.基于约束卷积神经网络的轿车款式识别[J].上海电力大学学报,2018,34(2):185-190.
JIA Zhentang.Vehicle Model Recognition Based on Constrained Convolutional Neural Network[J].Journal of Shanghai University of Electric Power,2018,34(2):185-190.
贾振堂.基于约束卷积神经网络的轿车款式识别[J].上海电力大学学报,2018,34(2):185-190.
JIA Zhentang.Vehicle Model Recognition Based on Constrained Convolutional Neural Network[J].Journal of Shanghai University of Electric Power,2018,34(2):185-190.