ATLAS (Automated Track & Lineside Asset Survey)

轨距测量是铁路网维护和运营的重要组成部分。轨距测量是Network Rail确保经过的列车与线路沿线物体和结构之间有足够的间隙的过程。目前，这一过程需要专业的测量机器，或通过手动审查从火车、无人机或人员操作的激光扫描仪捕获的点云数据。由MobiBiz(Lead)、BAE Systems和Jacobs Engineering Group组成的财团将加强对点云数据的解释。MobiBiz的基于云的ATLAS(自动跟踪和线边资产调查)平台将用于创建端到端的自动化管道，包括以下步骤：1.使用深度学习对点云中的轨道边结构进行自动分类、检测和分割并定位。这些结构包括道口、天桥、站台、下桥、隧道、高架桥、墙、信号、OLE支架、电信结构和其他线路侧设施。2.从点云数据中自动检测轨旁结构周围的植被。3.使用人工智能技术自动解释和分析点云，并辅助测量处理、铁路超高、曲率和间隙。4.提供基于Web的门户网站，以可视化使用点云数据和自动分析突出显示的地理参考路线，以帮助测量工程师进行可视化、验证和交叉检查。5.通过输出SC0文件和公开开放的API(应用程序接口)与现有的Network Rail系统集成，以促进与现有的报告系统、事件和资产数据库的集成。现在，车载捕获方法在捕获点云形式的海量信息方面极其高效和准确。然而，现在的问题在于处理。捕获TB级的数据会使处理和随后的分析变得极其缓慢，大部分过程需要将数据拆分成可管理的块、清理块并对数据进行分类。Atlas可以自动化这一部分，为Network Rail专业人员留出更多时间来实际审查输出的文件并做出明智的维护决策。这些检查可以更频繁地进行，在出现趋势和问题时发现它们。这些检查输出的经过处理的数据将用于通知维护，并协助进行预测性维护计划和干预。Atlas还提供了一个可视化平台，允许Network Rail工作人员比较多条路线上的结构的点云，以获取当前和历史数据。

Gauging is a vital part of Network Rail’s maintenance and operating of the railway. Gauging is the process through which Network Rail ensures adequate clearance between passing trains and lineside objects and structures. Currently, this process requires specialist gauging machines or via manual reviews of point cloud data captured from trains, drones or personnel operated laser scanners. The consortium of MobiBiz (Lead), BAE Systems and Jacobs Engineering Group will enhance the interpretation of point-cloud data. MobiBiz’s cloud based ATLAS (Automated Track & Lineside Asset Survey) platform will be used to create an end-to-end automated pipeline of the following steps: 1. Automatically classify, detect and segment and locate trackside structures within a point cloud using deep learning. These structures include crossings, overbridges, platforms, underbridges, tunnels, viaducts, wall, signals, OLE supports, telecommunications structures and other line side furniture. 2. Automatically detect vegetation around trackside structures from point cloud data. 3. Use Artificial Intelligence techniques to interpret and analyse point clouds automatically and assist in gauging processing, cant, curvature and clearance. 4. Provide a Web based portal to visualise geo-references route overplayed with point cloud data and automated analysis to assist gauging engineers for visualisation, validation and cross-checking. 5. Integrate with existing Network Rail systems by outputting SC0 files and exposing an open API (Application interface) to facilitate integration with existing reporting systems, events and asset databases. Trainborne capture methods are now extremely efficient and accurate in capturing vast amounts of information in the form of point clouds. However the problem now lies in processing. Capturing terabytes of data makes the processing and subsequent analysing extremely slow, with much of the process taken up splitting the data into manageable chunks, cleaning up the chunks and classifying the data. ATLAS can automate this part, leaving more time for specialist Network Rail staff to actually review the outputted files and make informed maintenance decisions. These inspections can be carried out more frequently, spotting trends and issues as they arise. The processed data output from these inspections will be used to inform maintenance and assist with predictive maintenance planning and intervention. ATLAS also provides a visualisation platform, allowing Network Rail staff to compare point clouds of structures across multiple routes for current and historical data.