Enhancing Infrastructure Resiliency Through Visual Data Analytics

最近发生的环境灾难——阿尔伯塔省的洪水、不列颠哥伦比亚省的野火和安大略省的风暴——的后果表明，我们在加拿大的建筑环境（建筑、桥梁、管道）远不能适应自然灾害。气候变化加剧了这些灾害；因此，找到改善社区安全和复原力的解决方案至关重要。拟议的研究使用基于自动视觉的方法来了解我们基础设施风险的性质，并加强我们对自然灾害的准备。通过先进的传感器和传感平台，可以以最小的资金和时间成本收集大量的视觉数据（图像和视频）。它们包含灾前和灾后基础设施的视觉和几何（2D/3D）以及时空（4D）信息。从这些数据中观察到的视觉变化提供了由于自然灾害造成的材料恶化或结构损坏（形状变形、开裂、倒塌）的明确迹象。从可视化数据中识别这些变化提供了一种评估风险和早期发现即将发生的基础设施中断的新手段。本研究的目标是通过将自动化集成到收集和分析视觉数据的方法中，提供计算算法来加速更安全、更有弹性的基础设施的开发。包括计算机视觉和深度学习算法在内的人工智能（AI）以及计算平台将以面向任务的方式利用，从基础设施中提取关键的视觉指标，这些指标将被捕获在相关的视觉数据中。从可视化数据中收集的信息将用于评估社区在自然灾害（洪水、飓风、地震）中的脆弱性和损失，监测其在环境和操作变化下的恶化情况，并实现数据驱动的调查，旨在填补阻碍现有建筑标准和设计规范的知识空白。这项研究将带来许多好处，解决加拿大和世界各地许多其他国家目前面临的基础设施挑战。该项目将使工程师和研究人员能够获得最有价值和最有用的视觉数据，以更好地评估我们的基础设施，并将为这些数据增加价值，以便将来重用和相关的科学研究。此外，学员将获得土木工程、人工智能和数据科学等严格的多学科技能，为他们未来在这些行业和学术界的就业做好准备。他们将在一个令人兴奋的新兴领域获得经验，为寻求利用计算机视觉方法和数据科学知识体系来解决广泛的土木工程问题的结构工程师提供新的机会。

The aftermath of recent environmental catastrophes - Alberta floods, British Columbia wildfires, and Ontario windstorms - reveals that our built environment (buildings, bridges, pipelines) in Canada is far from resilient to natural disasters. These disasters are exacerbated by climate change; therefore, it is critical to find solutions that improve community safety and resiliency. The proposed research uses an automated vision-based approach to understand the nature of the risk in our infrastructure and to enhance our preparedness for natural disasters. A large amount of visual data (images and videos) can be collected with minimal capital and time costs through advanced sensor and sensing platforms. They contain visual and geometric (2D/3D), and spatiotemporal (4D) information about the infrastructure before and after disaster events. Visual changes observed from these data provide clear indications of material deterioration or evidence of structural damage (shape deformation, cracking, collapsing) due to natural disasters. Identifying those changes from visual data provides a new means to assess risk and to enable early detection of impending infrastructure disruption. The goal of this research is to deliver the computational algorithm to accelerate the development of safer, more resilient infrastructure by integrating automation into the methods for collecting and analyzing visual data. Artificial intelligence (AI) encompassing computer vision and deep learning algorithms, and computing platforms will be leveraged in a task-oriented manner to extract critical visual indicators from infrastructure, which are captured in relevant visual data. The information collected from visual data will be used to assess communities' vulnerability and loss in natural disasters (flood, hurricane, earthquake), monitor their deteriorations under environmental and operational variations, and enable data-driven investigations that aim to fill the gaps in knowledge that hinder existing building standards and design codes. This research will deliver a number of benefits that address infrastructure challenges currently faced by Canadians and many others around the world. This program will empower engineers and researchers to capture the most valuable and useful visual data to better assess our infrastructure and will add value to that data for future reuse and relevant scientific study. Also, trainees will gain rigorous multidisciplinary skills in civil engineering, artificial intelligence, and data science, positioning them for future employment in these industries and academia. They will gain experience in an exciting emerging field, launching new opportunities for structural engineers seeking to exploit the body of knowledge in computer vision methods and data science to address a broad range of civil engineering problems.