摘要
地铁运营对轨行区界限具有非常高的要求,异物侵限对地铁运营会造成严重后果。现有研究方法大多为接触式或固定区域非接触式,这类方法无法对整条线路进行实时检测。针对该问题,提出了一种车载单目摄像机视野环境下异物侵限检测方法,即通过车载摄像机实时获取列车前方视野,利用改进的canny算法和霍夫直线检测算法确定轨行区形成感兴趣区域,并通过YOLO_v3深度神经网络实现对轨行区异物入侵检测。实验证明,该方法可有效提高检测效率和准确性。
Subway operation has a very high requirement for the boundaries of the track area,and foreign object intrusion can cause serious consequences.Most of the existing research methods are through contact detection or non-contact detection in fixed area,which cannot detect the whole track in real time.To address this problem,it proposes a method to detect foreign object intrusion through in-vehicle monocular camera,that is to capture the view in front of the train in real time through the in-vehicle camera,and use the improved Canny algorithm and the Hough linear detection algorithm to determine the track area and identify the area of interest,and realize the detection of foreign object intrusion in the track area by YOLO_v3 deep neural network.It is proved by experiments that this method can effectively improve the detection efficiency and accuracy.
作者
谭飞刚
Tan Feigang(School of Transportation and Environment,Shenzhen Institute of Information technology,Shenzhen,Guangdong,China 518172)
出处
《深圳信息职业技术学院学报》
2021年第5期71-76,共6页
Journal of Shenzhen Institute of Information Technology
基金
2021年广东省科技创新战略专项资金项目(项目编号:pdjh2021b0907)
广东省普通高校青年创新人才项目(项目编号:2020KQNCX205)。
关键词
车载单目摄像机
异物检测
侵限检测
地铁异物侵限
地铁轨行区
in-vehicle monocular camera
foreign object intrusion
intrusion detection
foreign object intrusion into subway track
subway track area
