基于SSD卷积神经网络的智能交通应用研究  

Research on the Application of Intelligent Transportation Based on SSD Convolutional Neural Networks

作　　者：金亚楠 严碧波[1] JIN Ya-nan;YAN Bi-bo(School of Electronic Information,Yangtze University,Jingzhou 434023,China)

机构地区：[1]长江大学电子信息学院

出　　处：《电脑知识与技术》2019年第7Z期211-213,共3页Computer Knowledge and Technology

摘　　要：针对传统的道路交通监控系统对目标的识别准确率较差、模型鲁棒性较弱且不具备实时性等问题,本文提出了一种基于卷积神经网络的道路交通目标检测方法。和传统的道路监控检测系统相比,卷积神经网络的模型更加适合处理大规模的数据。在卷积神经网络的分层特征提取下,既可以获得图像的低级语义,也可以获得图像的高级语义,经过对不同层级的样本特征分析,可以综合性地对目标进行识别分类,进一步提高模型的分类性能。验证结果表明该方法不但在准确率方面远远优于传统的识别系统,而且在硬件条件允许的情况下具有实时性。To address the problems that the traditional road traffic monitoring system has poor target recognition accuracy,weak model robustness and no real-time performance,a road traffic target recognition method based on convolutional neural networks is proposed.Compared with the traditional road monitoring and recognition system,the convolutional neural network model is more suitable for processing large-scale data.Under the hierarchical feature extraction of convolutional neural networks,both the lowlevel and the high-level semantics of the image can be obtained.After analyzing the sample features of different levels,the target can be comprehensively identified and classified,and the model’s classification performanc can be further improved as well.Ex?periments show that this method is not only superior to the traditional identification system in terms of accuracy,but also has realtime performance when the hardware conditions permit.

关 键 词：卷积神经网络 SSD模型 迁移学习 目标检测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]