Post-event structural damage assessment from images

在飓风或地震等自然灾害发生后立即对图像进行快速分类是应急管理和法医评估工作的关键。这些来自灾区的图像往往包括 ** 数千张图像，是宝贵的信息来源，可为有关资源分配和救灾工作优先次序的决策提供信息。虽然众包支持的计划，如2017年墨西哥中部地震后在墨西哥应用的计划，加快了整体评估过程，但众包数据可能不可靠或不完整，并可能导致评估结果之间的不一致。此外，用于查看图像的典型过程是手动且耗时的。加快和提高损失评估过程一致性的方法可以产生宝贵的回报，例如快速反应和确定保险索赔的优先次序。该提案旨在开发方法和软件 ** 工具，利用计算机视觉和机器 ** 学习技术简化事后结构损伤评估过程，以实现快速和自动处理大量图像。该 ** 项目的成果对行业合作伙伴既有商业价值，可以提供更好的和新的 ** 咨询服务，也有重大的社会价值，可以迅速应对洪水、暴风雪或地震等大规模灾害。

Rapidly classifying images immediately after a natural disaster such as a hurricane or an earthquake is key to**emergency management and forensic assessment efforts. Such images from disaster zones, which often include**thousands of images, are valuable sources of information that could inform decision-making regarding the**allotment of resources and prioritization of relief efforts. While crowdsourcing-supported schemes, like that**applied in Mexico following the 2017 Central Mexico earthquake, expedite the overall assessment process,**crowd-sourced data can be unreliable or incomplete, and can lead to inconsistencies between assessments**results. Additionally, the typical processes for reviewing images are manual and time-consuming. Methods for**accelerating and improving the consistency of the damage assessment process can yield invaluable returns such**as rapid response and prioritization of insurance claims. This proposal aims to develop methods and a software**tool to streamline the post-event structural damage assessment process using computer vision and machine**learning techniques to enable rapid and automated processing of large volumes of images. Outcomes of this**project have both commercial value to the industry partner in terms of enhanced and new offerings of**consulting services and significant societal value by being able to rapidly respond to large-scale disasters such**as floods, snowstorms or earthquakes.