摘要
为解决地铁列车前向运行环境中障碍物判断问题,提出基于视觉的地铁列车前向目标识别系统。该系统利用语义分割算法提取行驶轨道区域,结合轨道区域识别,基于SSD算法提取前向运行环境中轨道限界范围内的列车、行人等目标。针对上述场景,建立列车前向运行环境样本库,并基于该样本库完成模型训练。实验结果证明所得模型可有效识别列车前向行驶多目标。
In order to solve the problem of identifying obstacles in the metro train forward direction running environment, this paper proposes a visaul-based forward target recognition system. Taking into consideration of the track area recognition, and based on SSD algorithm to extract the train, pedestrian and other targets in the forward running environment, the system uses semantic segmentation algorithm to extract the running track area. In view of the above scenarios, it establishes the train forward operation environment sample database, and completes the model training based on the sample database. The experimental results show that the model effectively identifies the train moving forward multiple targets.
出处
《现代城市轨道交通》
2020年第1期86-89,共4页
Modern Urban Transit
关键词
地铁列车
视觉
深度学习
语义分割
目标识别
metro train
visual
deep learning
semantic segmentation
target recognition
