AUTOMATED SURVEY PROCESSING FOR RAILWAY STRUCTURE GAUGING

目前，轨道侧基础设施，如平台、高架桥、隧道等，都是使用网络铁路的激光测量系统进行扫描，并在存储在国家测量数据库（NGD）之前手动分类为结构类型。NGD存储有关铁路基础设施的信息，并用于计算列车与基础设施之间的间隙。目前，测量数据的处理是结构测量的一个主要瓶颈，收集的数据需要数年而不是数月才能以可用的格式进入数据库，用于测量计算。因此，许多数据都不在数据中，需要进一步的人工输入来管理许可，并确定管理新列车引入许可所需的行动。在某些情况下，这包括手动重新调查平台和其他结构以获取当前信息。这个Atkins研发项目旨在通过应用机器学习技术，利用我们在开发人工智能（特别是卷积神经网络）过程中的专业知识，自动对扫描数据进行分类，并识别任何植被，从而改善数据处理和可用性。基础设施数据将在3D走廊投影空间中考虑，而不是在2D部分中考虑，以提供更好的性能。这种方法将为数据提供更大的信心，从而更准确地理解结构与轨道之间的间隙。一旦该方法得到充分开发，处理数据的效率就会提高，从而实现这一点。这将使NGD拥有更多最新、一致和准确的调查数据。结果将减少调查数据处理时间，总体上降低了引进新列车或将列车级联到新线路的成本和时间表。SNC-Lavalin收购了Atkins集团公司，创建了一家全球全面整合的专业服务和项目管理公司。我们总共有5万多名员工。通过整合我们的员工队伍，我们为铁路和运输行业的客户提供了令人信服的服务。该项目将结合SNC-Lavalin的测量和基础设施团队以及Atkins软件和数据科学团队的技能，将所有关键领域的技术团队聚集在一起。

Currently the track side infrastructure such as platform, viaduct, tunnel etc. are scanned using laser measurement systems for Network Rail and categorized into structure types manually before storing in a National Gauging Database (NGD). The NGD stores information about the railway infrastructure and is used to calculate clearances between trains and infrastructure. Currently processing of survey data is a major bottle neck in structure gauging, with it taking years rather than months before data collected reaches the database in a useable format for gauging calculations. As a result much of the data is out of data in needs further human input to manage clearances and determine actions required to manage clearances for new train introductions. In some cases, this includes manual resurveying of platforms and other structures to obtain current information. This Atkins research and development project aims to improve data processing and availability by applying machine learning techniques using our expertise in developing Artificial Intelligence (Specifically Convolutional Neural Networks) processes to automatically classify the scanned data, as well as identifying any vegetation. The infrastructure data will be considered in a 3D Corridor Projection Space instead of 2D sections to provide better performance. This method will provide more greater confidence in the data leading to more accurate understanding of the clearances between structure and track. This will be achieved by the efficiency increases to process the data once the method is fully developed. This will lead to an NGD with more up to date, consistent and accurate survey data. A result there will be a reduction of survey data processing time with overall reduced costs and timescales for the introduction of new trains or cascading trains to new lines. SNC-Lavalin’s acquisition of the Atkins group of companies created a global fully integrated professional services and project management company. Together, we have over 50,000 employees. Through integrating our workforce we deliver a compelling offer to our clients in the Railway and Transportation sector. This project will combine the skills of SNC-Lavalin’s Gauging and Infrastructure teams and Atkins Software and Data Science teams, bringing together a technical team with all the key area.