Please select Into the mobile phone version | Continue to access the computer ver.

Building a Foreign Object Detection System 打造轨道异物入侵探测系统

[Copy Link]
Author: Livox Pioneer | Time: 2022-11-22 18:36:21 | Smart city|
0 6101








Rank: 9Rank: 9Rank: 9

Posted on 2022-11-22 18:36:21| All floors |Read mode
Railways are an important component of the Chinese transportation infrastructure and play a vital role in economic development. However, the country’s vast territory and complex and diverse terrain pose challenges to safe rail operations. Especially in the mountains of southwest China, railways are prone to hazards caused by foreign objects on the tracks, such as rockfalls. For example, on March 30, 2020, a passenger train from Jinan to Guangzhou was derailed because a landslide had affected the tracks, resulting in many injuries.

Figure 1 Train derailment on March 30 (Photo by Li Ga, a news reporter )

In addition to foreign objects on rail tracks caused by geological hazards, humans and animals can also wander onto train tracks.  The frequent intrusion of foreign objects on railway tracks severely affects the safe operation of railway systems. The timely and accurate detection of foreign objects and the triggering of corresponding alarms are crucial to railway safety.

Conventional foreign object intrusion monitoring systems for railway tracks feature dual power grids, optical fiber vibration sensing, and video monitoring. However, due to the lack of hardware capabilities, these systems cannot position foreign objects, their detection efficiency is low, and the false alarm rate and the missed detection rate are high. Lidar has been a success in the foreign object detection field due to its excellent detection capabilities and environmental adaptability. Livox assisted in the "Foreign Object Intrusion Detection Solution for Railway Tracks" solution. Livox provides a detailed introduction of the role of lidar in this solution. This solution has been implemented in the automated monitoring project for rockfall reduction for several railway lines.

Figure 2 Foreign object intrusion detection solution for Yuhuai Railway Lines: Integrates lidar, a camera, and a processing unit

Benefits of the Foreign Object Intrusion Detection Solution

The solution integrates various hardware devices, such as a lidar unit, a dome camera, a processing unit, and a headlight.

This solution allows you to obtain the 3D information of foreign objects, compensating for problems convention solutions have in detecting the size, depth, location, and other information of foreign objects. In addition, this solution does not depend on ambient light, avoiding some common issues with conventional lighting equipment. For example, conventional solutions attract moths, which can cause a high false alarm rate due to the limitations of visual-only solutions.

Figure 3 How Foreign Object Intrusion Detection on Railway Tracks Works

In this solution, the lidar is first used to extract the track surface from the scanned point cloud in the valid monitoring area. Inter-frame matching is performed for each frame when there are objects longer than 20 cm on the track surface. The differences between frames are analyzed to locate obstacles and calculate the distance between obstacles and the lidar. Then, the lidar reports the obstacle information to the dome camera and instructs the camera to rotate and focus on the foreign object and take a photo with flash. After the object is recognized from the image, the object is fed back to the monitoring center, and the personnel in the monitoring center checks whether the object is a foreign object.  In addition, the lidar can check whether the obstacle is static based on its distance to the object. In theory, only a static object that stays still for more than 15 seconds is recognized as a foreign object.

Figure 4 System on-site test

The Tele-15 is used as the main lidar, and Avia is used to compensate for the blind spots in different scenarios. Tele-15 supports long-range detection and highly dense point clouds. With 10% reflectivity, Tele-15 can measure objects at a distance of 320 m and even up to 500 m. The ultra-high point cloud density is adept at small object detection, allowing you to detect a pedestrian at a distance of 300 m and generate a point cloud with 12 points within 100 ms.

In roadside static scenarios, such as prewarning of foreign object intrusion, accurate prewarning is more important than the response time. The system leverages the unique non-repetitive scanning method of Livox lidar. As the scanning time increases, the number of points in the point cloud also increases. For example, 107 points can be detected for a pedestrian at a distance of 300 m within 1 second. Point clouds with higher density help the system accurately detect and recognize objects.

Figure 5 Left: Point cloud for a pedestrian at a distance of 300 m within 100 ms. Middle: Point cloud for a pedestrian at a distance of 300 m within 1 second. Right: Point cloud for a pedestrian at a distance of 100 m within 1 second

Below is a point cloud video of a single Tele-15 lidar monitoring the safety of a trackside section, which can still detect small objects as small as 20cm*20cm in size at a distance of 200 meters.

Accurate warning is the key

In addition, only a small amount of data needs to be transmitted in this solution, simplifying data transmission and analysis. The non-repetitive scanning capability provided by Livox Tele-15 and Avia greatly increases the monitoring efficiency in railway scenarios, and reduces the false alarm rate and the missed detection rate. After the solution was applied to Yuhuai Railway Lines, no false alarms or missed detection occurred, and normal objects on the railway tracks could be clearly distinguished from intruding foreign objects. No alarms are generated when a train passes a railway track without foreign objects. In comparison, when a member of the staff walked on the railway track, the foreign object intrusion detection system was triggered, transferring more accurate information to backend personnel.

Figure 6 Field installation diagram for more scenarios: Tunnel entrance/exit

This efficient and stable foreign object intrusion monitoring system was developed by integrating the advantages of lidar and conventional cameras. The system improves the accuracy of foreign object intrusion detection, making it of great social and economic significance in ensuring safe railway operations. Moreover, Livox hopes to continuously release cost-effective lidar products to reduce the overall cost of the solution. This way, the state-of-the-art solution can be applied to more miles of track and prevent more accidents.


图1 3月30日拍摄的火车脱轨现场(无人机照片)新华社记者 李尕 摄


传统铁路异物入侵监测系统主要以双电网、光纤振动、视频监控等手段为主,当前受制于系统硬件能力的不足,在实际使用中存在探测效率低、无法定位异物、误报率和漏报率偏高等问题。而激光雷达凭借其优异的探测能力和环境适应性,在异物探测领域崭露头角。下面通过某客户的「轨道异物入侵探测方案」, 详细介绍激光雷达在其中扮演的角色。该方案已经在某个危岩整治自动化监测项目中实际落地。

图2 轨道旁异物入侵探测方案现场图:融合了激光雷达、摄像头、处理单元



在实际操作中,系统先通过激光雷达对有效监控区间内的扫描点云进行轨面提取,对轨面上尺寸大于20cm 的物体进行每一帧的帧间匹配,分析每一帧的差异定位到障碍物及障碍物相对激光雷达的距离。然后激光雷达将障碍物信息反馈给球机,指导球机转动并对焦到异物的位置进行闪光拍照。通过图像进行目标识别后反馈给监控中心,由监控中心工作人员判断是否属于异物。且雷达可以通过目标距离判断出障碍物是否是静态,理论上只有连续15s以上的静态目标才被识别为异物。



泰览 Tele-15 作为主激光雷达,负责超远距离的探测;根据不同场景,也可辅助使用 Avia 或 HAP 激光雷达进行水平方向覆盖。Tele-15 具有量程超远、点云密度超高的特性,10%反射率即可测距320m,最远可探测500m外物体;超高点云密度对于小物体探测具有天然的优势,行人在300m处依旧可以探测到12个点(100ms时间内);

左:300米处行人点云(100ms) 中:300米处行人(1s) 右:100米处行人(1s)

下面是单个Tele-15 激光雷达对铁轨旁路段进行安全监控的点云视频,实测200米开外仍可稳定探测到尺寸仅为 20cm*20cm 的小物体。


在异物入侵预警这类路侧静态场景,对响应时间要求其实不高,关键是要准确预警。系统充分发挥了 Livox 雷达独有的非重复扫描方式,随着扫描时间增加,点云数量也会增加,例如行人在相同距离300m处可以探测107个点(1秒扫描时间)。更高的点云密度意味着客户能更好地探测识别到异常物体。

同时,该方案执行过程中数据传输量小,有效降低了数据传输及分析的难度;而 Livox Tele-15 和 Avia、HAP 的非重复扫描也能极大拓展铁轨应用场景下的监测效率,降低系统的误报率和漏报率。方案可清晰区分出铁路上的正常物体与闯入的异物;正常行驶通过的列车不报警,而在铁路上的工作人员则会触发异物入侵探测系统,将最准确真实的信息传递给后台工作人员。

图3 更多场景下的现场安装图: 隧道出入口

充分融合激光雷达与传统相机两者自身的优势,研发出高效、稳定的异物入侵监测系统,提高异物入侵探测准确性,这对于保障铁路的安全运营,具有重大的社会意义和经济价值。Livox 也希望通过不断推出高性价比的激光雷达,降低整个方案的成本,让这种新型的解决方案得以在更大范围应用中落地,守卫更多宝贵的生命。

This post contains more resources

You need to Login Before they can download or view these resources, Don’t have an account?Register


Use props Report

You need to log in before you can reply Login | Register

Credit Rules

Quick Reply Back to top