摘要
随着地铁列车无人驾驶技术的发展,有必要增加线路障碍物监测系统功能,对列车运行前方障碍物进行主动探测、识别和预警。列车前方轨道识别及构建限界空间是列车车前障碍物监测的关键技术之一,只有在准确识别出轨道限界的条件下才能够判断列车前方检测到的目标的危险性。为解决单个传感器进行轨道识别存在的缺陷,提出一种融合可见光相机和激光雷达传感器,并借助深度卷积神经网络和点云特征,实现轨道识别的新方法。
With the development of unmanned driving technology in metro transportation systems,the function of obstacle detection,which means actively detecting and identifying the intrusions in front of the train and sending alarms,has become necessary.One of the key technologies of obstacle detection is to identify the track ahead of the train and to construct the boundary of the track area.With an accurately identified track boundary,the danger level of the detected target in front of the train can be precisely judged.In order to solve the defects of single sensors for rail track recognition,a new method is proposed by combining optical camera and LiDAR,using deep convolutional neural network and point cloud features to recognize rail track area.
作者
朵建华
杨柏钟
Duo Jianhua;Yang Baizhong
出处
《现代城市轨道交通》
2021年第9期98-102,共5页
Modern Urban Transit
基金
浙江省重点研发计划《轨道交通核心关键器件及产品研发-城市轨道交通自主驾驶DTO列车运行控制系统关键技术研究》(2019C01144)。
关键词
地铁
轨道识别
多传感器
卷积神经网络
metro
rail recognition
multi-sensor
convolutional neural network