Full automation of sewer CCTV surveys

英国（和世界各地）的水务公司定期检查其下水道，以优先进行维护，并确保其网络的有效运行。如果不这样做，可能会导致事故，包括未经处理的污水排放到环境中，管道倒塌，甚至形成下水道堵塞的脂肪山。UKWIR的目标是到2050年实现无控制的下水道排放为零，这加强了减少这些事件的重要性。在大多数情况下，这些事件可以通过闭路电视测量来预防，并通过早期干预来解决。然而，调查既费时又昂贵。此外，这些报告往往是不一致和不准确的，主要是由于人为错误和故障代码的主观性质。该项目旨在增强现有的注释和报告过程，总体目标是实现CCTV调查过程的完全自动化。人工智能和机器人技术的结合将彻底改变下水道的测量和维护，提高整个实践的速度、准确性和效率。反过来，这将导致完成更多的调查和更高的机会，先发制人的下水道故障。目前，SWW和UOE正在完成一个KTP项目，内部实施原型故障检测方法，在前面的博士研究。这项为期两年的合作伙伴关系（将于2020年11月完成）已开发并培训了SWW闭路电视录像档案的检测系统，并将其作为决策支持工具。这能够突出故障并从记录的CCTV镜头中估计其一般类型;对于快速分析以前未使用的缺乏注释的视频非常有用。除了技术开发，该项目还建立了一个合作者网络（包括iTouch和WRC），同时在学术和行业活动中广泛宣传。虽然KTP已经实现了为SWW带来功能工具的目标，但很明显，该技术具有更大的潜力，可以提高当前实践的效率和准确性。该项目的三个主要目标是：（1）开发该技术的注释功能，以实现MSCC中概述的完整标准。(2)实施开发的软件，以协助和执行现场报告。(3)记录和注释以前未报告的管道功能。拟议的项目提供了机会，不仅发展这项研究成为一个充分蓬勃发展的技术，为英国和国际使用，但提供了资源和基础，为未来的图像处理和机器学习研究在SWW和整个水行业。这项研究将继续为国家和全球倡议提供解决方案，与联合国可持续发展目标（“保护陆地和淡水生物多样性的重要场所”），UKWIR的大问题（“我们如何在2050年前实现下水道零排放？”）保持一致。”）和英国工业战略（“利用人工智能提高部门生产力”）。无论是未来的目视检查技术还是自动化和其他操作功能的支持，这项工作都将继续使用尖端的计算机科学技术来提高效率和改善性能。

Water companies across the UK (and world) regularly inspect their sewers to prioritise maintenance and ensure the effective operation of their network. Failure to do so can result in incidents, including the discharge of untreated sewage to the environment, pipe collapse or even the formation of sewer blocking fatbergs. The importance of minimising these events is reinforced by the UKWIR objective to achieve zero uncontrolled sewer discharges by 2050. In most cases these occurrences are prevented using CCTV surveying and resolved with an early intervention. However, surveys are time consuming and expensive. Moreover, these reports are often inconsistent and inaccurate, largely due to human error and the subjective nature of fault codes. This project aims to augment the existing annotation and reporting process, with the overall ambition of fully automating the full CCTV surveying process. This proposed combination of AI and robotics will revolutionise sewer surveying and maintenance, improving the speed accuracy and efficiency of the entire practice. In turn this should result in the completion of more surveys and a much higher chance of pre-empting sewer failure.Currently SWW and the UoE are completing a KTP project, to internally implement the prototype fault detection method, investigated during the preceding PhD. The two-year partnership (due to complete in November 2020), has developed and trained the detection system on SWW's archive of CCTV footage and implementing this as a decision support tool. This is capable of highlighting faults and estimating their general type from recorded CCTV footage; extremely useful for the quick analysis of previously unused video that lacks annotation. Alongside technical developments, the project has built a network of collaborators (including iTouch and the WRc), whilst being widely publicised at both academic and industry events. Although the KTP has achieved its goal of bringing a functional tool to SWW, it is clear that the technology has potential for so much more, driving up efficiency and accuracy over current practices. The three key goals of the project are:(1) Develop the annotation capabilities of the technology to achieve the full standards outlined in the MSCC.(2) Implement the developed software so as to assist and perform live reporting.(3) Record and annotate previously unreported pipe features.The proposed project offers the opportunity to not only develop this research into a fully flourished technology for both UK and international use, but provides the resources and foundations for future image processing and machine learning research within SWW and the water industry as a whole. This research would continue to contribute solutions to national and global initiatives, aligning with the UN sustainable development goal ('protecting important sites for terrestrial and freshwater biodiversity'), UKWIR's Big Questions ('How do we achieve zero uncontrolled discharges from sewers by 2050?') and the UK industrial Strategy ('Increase sector productivity utilising AI'). Whether this takes the form of future visual inspection techniques or automation and support of other operational functions, the work would continue to drive efficiencies and improve performance using cutting edge computer science techniques.