Machine learning enabled remote infrastructure inspection tool using both existing and new imagery

使用现有和新图像的机器学习支持远程基础设施检查工具

• 批准号: 71546
• 金额: $ 6.23万
• 依托单位: MAESTRANO GROUP PLC
• 依托单位国家: 英国
• 项目类别: Feasibility Studies
• 财政年份: 2020
• 资助国家: 英国
• 起止时间: 2020 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=71546
• 关键词: Machine; learning; enabled; remote; infrastructure

Infrastructure in the built environment (buildings, roads, rail, energy structures and more) requires inspection regularly, for safety. New technologies such as drones, mobile phone cameras, CCTV and digital cameras mean that there is more and more imaging data of assets available, making it ever increasingly difficult for asset inspectors and managers to store, review and assess captured data. But there is too much data to review and so costly (and at times) physical inspections still occur, and that will be increasingly difficult in a post-pandemic world of restricted movement and both public and private budget constraints.The key objective for the project is to develop a fully standardised imagery and video data review system to deliver the results in a cloud-based user interface on top of an already developed and powerful ML technology. This is innovative as for the first time it will enable asset managers and inspectors to review very high volumes of inspection data from remote systems such as drones and on-vehicle cameras on an exception basis via ML based automatic recommendation of candidate issues and problems, and from any internet-connected personal computer, in home office or at any location as required.

建筑环境中的基础设施（建筑物、道路、铁路、能源结构等）需要定期检查以确保安全。无人机、手机摄像头、闭路电视和数码相机等新技术意味着可用的资产成像数据越来越多，使得资产检查人员和管理人员存储、审查和评估捕获的数据变得越来越困难。但需要审查的数据太多，因此成本高昂（有时）的实物检查仍然会发生，在疫情后流动受限、公共和私人预算都受到限制的世界中，这将变得越来越困难。该项目的主要目标的目标是开发一个完全标准化的图像和视频数据审查系统，在已经开发的强大机器学习技术之上，在基于云的用户界面中提供结果。这是创新性的，因为它将首次使资产管理者和检查员能够通过基于机器学习的候选问题自动推荐，在例外情况下审查来自无人机和车载摄像头等远程系统的大量检查数据，以及从家庭办公室或任何需要的位置的任何连接互联网的个人计算机。