防护墙和地面振动光纤与视频配对数据

本数据集是一个面向铁路周界入侵检测的多模态融合感知数据集，综合了基于相敏光时域反射仪（Φ-OTDR）系统采集的光纤振动数据与视频监控技术采集的图像数据。数据集通过同步采集振动信号与视频图像，形成由文件名时间信息对应组合的数据对，覆盖人员敲击防护栏、攀爬防护栏及行走三类典型入侵场景，共包含598个数据对文件。研究旨在突破传统单模态传感在复杂环境下受噪声干扰大、误报率高的技术瓶颈，通过多模态数据融合提升目标识别的鲁棒性，为铁路周界入侵检测算法的开发与验证提供关键数据支撑。 在雄安铁路线附近的防护墙和地面现场场景中采集。采集设备采用桂林光翼DAS一体机，相机神戎高清激光夜视仪。振动光纤一段布置在防护墙上，另一段浅埋在地面，相机安装在防护墙附近。振动光纤检测铁路附近环境中的振动信号并对异常振动进行报警，相机采集和检测异常振动位置的视频数据，确认异常振动的威胁性。通过振动光纤与视频联动检测，提高铁路周界入侵事件的识别与报警的准确率和效率，从而显著提升了列车运行环境的安全性。数据量10.6GB。This dataset is a multimodal fusion perception dataset for railway perimeter intrusion detection, integrating fiber optic vibration data collected by a phase-sensitive optical time-domain reflectometer (Φ-OTDR) system and image data captured via video surveillance technology. The dataset synchronously collects vibration signals and video images to form data pairs matched by filename timestamp information, covering three typical intrusion scenarios: personnel striking guardrails, climbing guardrails, and walking, with a total of 598 data pair files. This study aims to break through the technical bottlenecks of traditional single-modal sensing, such as high noise interference and elevated false alarm rates in complex environments, improve the robustness of target recognition through multimodal data fusion, and provide critical data support for the development and verification of railway perimeter intrusion detection algorithms. Data was collected at protective wall and on-ground field sites near the Xiong'an Railway Line. The collection equipment includes the Guilin Guangyi DAS all-in-one machine and the Shenrong high-definition laser night vision camera. One segment of the vibration fiber was laid on the protective wall, while the other was shallowly buried in the ground, and the camera was installed near the protective wall. The vibration fiber detects vibration signals in the environment near the railway and triggers alarms for abnormal vibrations, while the camera collects and inspects video data at the abnormal vibration location to confirm the threat posed by the abnormal vibration. Through coordinated detection of vibration fiber and video, the accuracy and efficiency of railway perimeter intrusion event recognition and alarm are improved, thereby significantly enhancing the safety of train operation environments. The total data volume is 10.6 GB.

搜集汇总

背景与挑战

背景概述

该数据集为面向铁路周界入侵检测的多模态融合感知数据集，同步采集光纤振动信号与视频图像，形成了598个数据对，涵盖人员敲击、攀爬防护栏及行走三类典型场景。其旨在通过多模态数据融合提升复杂环境下目标识别的准确性与鲁棒性，为相关算法开发提供关键数据支撑。