IAEA Symposium on International Safeguards

Leveraging Computer Vision for Imagery Analytics

Not scheduled

5m

M Building (Vienna International Centre)

Roundtable Member Leveraging technological advancements for safeguards applications (TEC) [TEC] Collection, Processing and Analysis of Satellite and Open Source Imagery Data

Speaker

Catherine Dill (Middlebury Institute of International Studies at Monterey)

Description

This paper will explore the potential applications and limits of computer vision for analyzing satellite imagery and ground imagery for safeguards purposes. The paper will first detail the current state of the art as it relates to computer vision for three main tasks related to ground and satellite imagery analysis: land classification, change detection, and object identification. The paper will consider the availability of both ground and satellite imagery, and in particular, the effect of higher revisit rates from small satellites on machine learning prospects. Under the banner of computer vision, this paper will then explore the available toolkits and machine learning algorithmic approaches to these tasks that might be useful in a future context. This paper will then examine how these techniques may be leveraged for informing safeguards assessments at the IAEA. Finally, the paper will evaluate both the technical and practical limitations of incorporating computer vision-based imagery analytics into safeguards.