文章基本信息

标题：Deep Learning Based Vehicle Tracking in Traffic Management
本地全文：下载
作者：Mohamed Shehata ; Reda Abo-Alez ; Farid Zaghlool 等
期刊名称：International Journal of Computer Trends and Technology
电子版ISSN：2231-2803
出版年度：2019
卷号：67
期号：3
页码：5-8
DOI：10.14445/22312803/IJCTT-V67I3P102
出版社：Seventh Sense Research Group
摘要：Nowadays the visual vehicle tracking system (VTS) becomes a vital part of the Intelligent Transportation System (ITS). It is the cornerstone of vehicles behavior analysis. Our methodology for developing a VTS achieves video based vehicles detection, classification and tracking. In the detection process a deep machine learning system based on faster region conventional neural network (FasterRCNN) detector is used. In the classification process a deep machine learning system based on conventional neural network (CNN) is used. In the tracking process, motion vector estimation (MVE) algorithm is used to determine vehicles directions and positions in the video frames. Finally vehicle behaviour understanding algorithm based on vehicle trajectory implementation and vehicle speed calculation is used to manage the traffic flow. After testing the developed VTS, the results show that, 95% of the tested vehicles are precisely detected, 90% of the detected vehicles are successfully classified, and 92% of detected vehicles tracks are well generated.
关键词：Video Processing; Faster-RCNN; CNN; Motion Vector Estimation.