基于YOLO v3的有轨电车在途障碍物检测方法

YOLO v3-based Obstacle Detection Method for Tram in Transit

有轨电车一般使用混合路权,与社会车辆接触密切,安全性需进一步提升。以视觉传感器与YOLO v3卷积神经网络为基础,通过对YOLO v3网络模型损失函数的优化,提出一种准确率较高的有轨电车在途障碍物识别方法,实时识别列车运行图中可能遇到的各种危险因素,辅助司机驾驶,提升车辆运行安全性。经过验证,本方法具有较高的准确性和鲁棒性,是一种有效的列车驾驶辅助手段。

Trams generally use mixed right-of-way and are in close contact with automobiles,safety still needs to be further improved.Based on the vision sensor and YOLO v3 convolutional neural network,this paper proposes a method to identify the obstacles of a tram in transit with high accuracy by optimizing the loss function of YOLO v3 network model,and carries out the identifi cation of various risk factors that may be encountered in the train diagram to assist the driver and improve operation safety.It is proved that this method has high accuracy and robustness,and is an eff ective train driving aid means.