Use of Artificial Intelligence towards Automation of Analog Seismogram Digitization

自19世纪后期以来，地震仪一直在记录地面运动，大约100年来的记录都是模拟形式（例如，纸质记录），无法使用现代技术进行检查。这些数据包含地震、火山爆发、地下变化、天气模式变化等信息，对于理解影响地球的各种现象以及它们如何随时间演变至关重要。该项目旨在通过将人工智能引入转换过程，在将这些模拟数据转换为可用的数字格式方面迈出重要一步。目前，有一种软件可以记录图像并生成数字地震图，但它需要大量的人工交互，使得这个过程缓慢、不切实际或不可能。人工智能的成功实施将使更多的数据得到快速处理，供科学界使用，这是一个重要的更广泛的影响。该项目将从检查数字化分析开始，以确定和建立用于构建神经网络的训练数据库。研究人员还将确定将从人工智能程序的实施中获益最多的步骤，以减少人类互动并提高地震记录数字化的准确性。用于图像分类和目标识别的神经网络现已可用，并将进行研究，以找到最适合地震记录数字化过程的算法。改进后的数字化软件将公开提供，以增加用户，并提供一个强大的工具，将模拟地震图图像转换为研究质量的数字地震图。它将使地震学界能够检索用于现代分析的数据，并为进行新型研究提供机会。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Seismometers have been recording ground motion since the late 1800s, and about 100 years of the recordings are in analog form (e.g., paper recordings) that cannot be examined using modern techniques. These data contain information about earthquakes, volcano eruptions, subsurface changes, changes in weather pattern, to name the few, and are vital for understanding various phenomena that affect the Earth and how they evolve over time. This project aims to make a significant step forward in converting these analog data into usable digital format by introducing artificial intelligence to the conversion process. Currently, there is a software that takes a record image and generates digital seismograms, but it requires substantial human interaction making this process slow, impractical, or impossible. Successful implementation of artificial intelligence will allow more data to be processed quickly for use by the scientific community, which is a significant broader impact. The project will start by examining the digitized analyses to determine and build the training database to be used for the construction of the neural network. The investigators will also identify steps that will benefit most from implementation of artificial intelligence procedures to decrease human interaction and improve accuracy of the digitization of seismograms. Neural networks for image classification and object identification are now available and will be examined to find the algorithm that is most suitable for the seismogram digitization process. The improved digitization software will be openly available to increase users and provide a robust tool to convert analog seismogram images to research-quality digital seismograms. It will enable the seismological community to retrieve data that for application of modern analyses, and open opportunities for new types of research to be done.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.