Modern approaches to extracting precise measurements of structure from images rely on multiple images from different vantage points (Structure from Motion, or SfM) or from known relationships between the camera orientation and the scene being imaged. But what happens when neither of these conditions are met? What if a disaster recovery team needs to assess the damage of a structure? A 3D model of the building is of immense value. But the team has only one image - drawn from a social media post - of the damaged structure. Traditional methods cannot be used to extract measurements from this one image. Projective geometry and perspective have long been used by artists to recreate scenes as seen from the eye or from a camera using vanishing points. These same vanishing points can be used in reverse to recreate the 3D structure and associated camera model in single images. While this has been used for many years in the photogrammetric community, error propagation - specifying the confidence intervals around measurements - in these contexts is critically important and difficult to get right. 

Dr. Settergren will introduce vanishing point analysis and discuss various means of applying rigorous error propagation techniques in the analysis. He will also discuss the history ofBAE Systems and opportunities for students at the company.

Dr. Reuben Settergren is a software developer and photogrammetrist at BAE Systems. For the last twenty years he has contributed to the SOCET GXP application - an advanced image processing and photogrammetric tool suite for extracting information from geospatial images. Since Fall 2020, Reuben has also been teaching a Computer Science elective for high school seniors, at The Cambridge School in San Diego. He received his Bachelor of Science at Johns Hopkins, earned a PhD in Operations Research from Rutgers, taught Math and Computer Science at the College of Wooster, and did a postdoc at Imperial College, London. 

This lecture is sponsored by the Lecia Geosystems Chair in Geospatial Imaging. The Leica Geosystems-endowed fund was established in 2005. Their gift-in-kind created the Leica Geosystems Chair in Geospatial Imaging at CAST providing advanced photogrammetric and remote-sensing software, maintenance, support, and services to the Center. This endowment also serves to build an academic discipline in geospatial modeling and expand resources for analyzing, storing, and communicating geospatial research data.