Patrick O'Neil
Establishing Pattern-of-Life from Rapid Revisit Satellite Imagery using Machine Learning

Wednesday, April 10, 2019
Building 3 auditorium -11:00 AM
(Cookies at 10:30 AM)

The last decade has seen an explosion in the availability and affordability of commercial satellite imagery. This growth has yielded tremendous improvements in our ability to perform environmental monitoring, commercial development, and defense and intelligence planning. Recently, BlackSky has launched the first of its Globals constellation of imaging small sats. With revisit rates of up to eight times a day, this constellation will offer temporal resolution unattainable by current commercial satellite imaging solutions. While this level of insight enables many critical new use cases, the deluge of data has made it difficult for imagery analysts to process all the data and prioritize their efforts. This presents an immediate need for novel machine learning and computer vision techniques which can identify and flag significant changes among thousands of images per day. Through the use of deep learning, we are able to extract objects from scenes and establish temporal baselines of activity at facilities. By continually monitoring these facilities, we can establish pattern-of-life and, importantly, any deviations from the established patterns. In this talk, I will describe the machine learning system we have built at BlackSky to process, analyze, and enrich the tremendous volume of data we capture from our satellites every day.

Patrick O'Neil is the Director of Machine Learning and Artificial Intelligence at Spaceflight Industries and an affiliate faculty member at George Mason University. In his role at Spaceflight, he directs machine learning and artificial intelligence research and development efforts across the company. At George Mason, he offers a weekly seminar focusing on applying deep learning methods to satellite imagery. Patrick has spent his career in the satellite imagery industry, working for companies such as Spadac, GeoEye, and OpenWhere. In 2017, Patrick received his Ph.D. in Mathematics from George Mason University where his research concerned analyzing point clouds using topological data analysis methods. Prior to his Ph.D. work, Patrick received his Bachelor's degree from Virginia Tech, also studying mathematics.

IS&T Colloquium Committee Host: Dr. Carlos Cruz

Sign language interpreter upon request: 301-286-7348

Request future announcements: ist-colloq-subscribe@lists.nasa.gov