EAGER: Active Citizen Engagement to Enable Lifecycle Management of Infrastructure Systems

结构中的缺陷和缺陷在结构的使用寿命内逐渐演变，必须定期评估潜在的退化，以做出合理的管理决策并确定修复优先级。人类观察仍然是此类评估的主要机制，既耗时又昂贵。公民科学和众包提供了从多个角度、以频繁的时间间隔、在多种条件下收集特定结构的大量照片的机会。这些视觉数据跨越时间和空间，可以详细记录随时间推移的恶化情况。这个早期概念探索性研究资助 (EAGER) 项目将利用计算机视觉领域的最新知识，自动执行与生命周期结构评估和管理相关的许多任务，从而降低生命周期成本和风险。该项目开发的方法还将为寻求利用数据科学解决广泛的结构工程问题的结构工程师带来新的机会。将开发一个展示我们目标结构方法的流媒体视频，以供广泛传播和学生参与。来自未受过工程师培训的公民的日常图像与工程师捕获的图像类型有很大不同。这些图像的大部分包含与工程目的无关的信息，并且这些图像的自动处理会产生错误的结论。此外，这些图像是从随机位置和角度收集的，缺乏比例和方向信息。然而，这些障碍是可以克服的。该项目将纳入有关结构评估过程的基本知识，以便将这些图像用于工程目的。每个查询图像和目标结构模型之间的几何关系将通过匹配它们的局部特征来计算。将执行这些图像中的每一个相对于目标结构的自动定位，并且将提取图像的相关部分（称为感兴趣区域）以供人或机器进行结构评估。将使用通过社交媒体从积极参与的公民收集的图像进行实验验证。将针对结构中特别脆弱的区域对该方法的能力进行定量评估。该项目将克服使用公民科学家的视觉数据的固有挑战，促进我们执行生命周期结构评估方式的转变。

Flaws and defects in structures evolve gradually over a structure's lifetime, and potential degradation must be evaluated periodically to make sound management decisions and set repair priorities. Human observation, still the predominant mechanism for such evaluation, is time-consuming and costly. Citizen science and crowdsourcing provide opportunities to collect large numbers of photos of certain structures, from many perspectives, at frequent intervals, and under many conditions. Such visual data reach across both time and space, enabling a detailed record of deterioration over time. This EArly-concept Grant for Exploratory Research (EAGER) project will exploit the latest knowledge in computer vision to automate many tasks related to lifecycle structural evaluation and management, reducing both lifecycle cost and risk. The methodology developed within this project will also launch new opportunities for structural engineers seeking to exploit data science to address a broad range of structural engineering problems. A streaming video demonstrating the methodology on our target structure will be developed for broad dissemination and student engagement. Everyday images from citizens, not trained as engineers, are very different than the types of images than engineers capture. Large portions of these images have information irrelevant for engineering purposes and automated processing of these images would generate faulty conclusions. Furthermore, these images are collected from random locations and perspectives, and lack scale and orientation information. However, these barriers can be overcome. This project will incorporate essential knowledge about the structural evaluation process to enable the use of these images for engineering purposes. Geometric relationships between each of the query images and the model of the target structure will be computed by matching their local features. Automatic localization of each of these images with respect to the target structure will be performed, and relevant portion of the images, called the region of interest, will be extracted for structural evaluation by human or machine. Experimental validation will be performed using images collected from active engaged citizens through social media. A quantitative evaluation of the capabilities of the methodology will be performed, targeting especially vulnerable regions of a structure. This project will overcome the inherent challenges in using visual data from citizen scientists, facilitating a transformation in how we perform lifecycle structural evaluation.