摘要
设计和实现了一种基于目标检测和多目标跟踪的“双轮车危险驾驶行为智能管控系统”。重点研发目标检测技术,具体包括基于YOLOv4-tiny骨干的注意力网络以提高目标检测精度,和多尺度空洞卷积网络以适应不同大小目标的检测,以提高目标识别与定位的准确性。系统管理模块采用数据可视化交互页面以实现智能管控,可广泛部署在交通主要路口,让交警执法有的放矢。
In this paper, an "intelligent control system for dangerous driving behavior of two-wheeled vehicle" based on target detection and multi-target tracking is designed and implemented. The system focuses on the research and development of target detection technology, including the attention network based on YOLOv4-tiny backbone to improve the target detection accuracy, and the multi-scale atrous convolution network to adapt to the detection of targets of different sizes to improve the accuracy of target recognition and positioning. The system management module realizes the intelligent management and control through the data visualization interaction page, which can be widely deployed at the main traffic intersections, and allow the traffic police to enforce the law with a definite object in view.
作者
林静怡
史晓颖
Lin Jingyi;Shi Xiaoying(Computer&Software School,Hangzhou Dianzi University,Hangzhou,Zhejiang 310018,China)
出处
《计算机时代》
2022年第10期64-68,共5页
Computer Era
关键词
目标检测
多目标跟踪
可视化交互
智能管控
object detection
multiple target tracking
visual interaction
intelligent control
