基于快速背景差分的高速铁路异物侵入检测算法

随着我国高速铁路通车里程不断增加,高速铁路的运营安全备受关注,异物侵入铁路限界对运营安全危害极大,有效检测侵入线路净空的异物对保障高速铁路安全运营具有重要意义。铁路场景环境光线多变和图像通道众多的特点对基于图像的异物检测方法的处理效果和实时性提出了较高的要求。针对铁路场景抖动发生在垂直方向的特点,提出了一维灰度投影结合高斯滤波的图像快速去抖方法,在大幅提高处理速度的同时获得了较好的去抖效果;针对复杂多变的背景,提出了一种基于前景目标统计分布的背景更新算法,定义了目标分散指数用于确定行列投影次序,通过统计前景目标分布实现背景更新,在提高速度的同时解决了传统背景更新算法难以解决的鬼影问题。最后通过背景差分获取前景目标,并通过目标标记、合并与特性分析提高目标检测的准确性。沪宁城际高速铁路典型场景的现场实验表明,该算法能有效检出铁路场景侵限目标,系统综合误检率约为0.54%,漏检率为0。

Highspeed Railway Clearance Intrusion Detection Algorithm with Fast Background Subtraction

With the rapid development of high-speed railway, the railway safe operation attracts more and more attention. Foreign objects intruding the railway clearance is a great threat to normal operation. An effective intruding detection method is important to high-speed railway safety. The variable ambient light and the large amount channels of image need an efficient and effective algorithm. After the analysis of vertical jitter direction, a fast image stabilization algorithm based on one-dimensional gray projection matching combined with Gauss filtering is proposed to accelerate the speed and improve the effect. According to the complex and ever-changing background, a fast background updating algorithm based the foreground object distribution is proposed. The target dispersion index is defined to determine the projection order of horizontal and vertical direction. This algorithm not only improves the updating speed but also solves the ghost problem. The system using this algorithm has been build and tested on Shanghai-Nanjing Intercity High-speed Railway line. The experiment in typical railway scenes shows that the system can detect intruding objects under most whether condition. The false alarm rate is about 0.54%, and the missing rate is 0.