车载单目摄像机下地铁轨行区检测与提取

Detection and Extraction of Rail Transit Area under Vehicle Mounted Monocular Camera

地铁对线路运行区域(即轨行区)有非常严格的要求,异物侵入轨行区检测可有效提醒司机避免行车事故发生。轨行区域检测与提取作为轨行区异物自动检测中第一步具有重要的研究意义,在分析现有研究方法的基础上提出一种改进的轨行区检测算法。首先,对图像进行预处理完后利用canny算子进行边缘检测;然后,利用霍夫直线检测算法构建钢轨直线候选集;再次,利用直线斜率、直线端点偏移量等约束对钢轨直线候选集进行处理,移除异常数据并有效合并直线段;最后,用钢轨直线端点提取轨行区,为后续轨行区异物自动检测提供坚实基础。对某市地铁运行线路真实数据进行采集及实验,结果表明,本文所提方法可准确检测出地面区间、地面站台、隧道及地下车站等场景的轨行区,该方法提高了钢轨检测准确性和轨行区提取准确性。

Rail transit has very strict requirements on the line operation area(i.e.rail transit area).The detection of foreign matter intrusion into the rail transit area can effectively remind drivers to avoid traffic accidents.As the first step in the automatic detection of foreign objects in the track area,track area detection and extraction bears very important research significance.Based on the analysis of existing research methods,an improved track area detection algorithm is proposed in this paper.Firstly,Canny operator is used for edge detection after image preprocessing,and then Hough line detection algorithm is applied to construct rail line candidate set.In addition,constraints such as line slope and line endpoint offset are employed to process rail line candidate set to remove abnormal data and effectively merge line segments.Finally,the rail straight line endpoint is needed to extract rail line area,and it provides a solid foundation for the automatic detection of foreign matters in the track area.The real data collection and experiment of a metro line show that the proposed method can accurately detect the track area of the ground section,ground platform,tunnel and underground station,and improve the accuracy of rail detection and track area extraction.