Overcoming Challenges in 3D Reconstruction and Localization: Using the Resulting Geometry for Material Recognition

Overcoming Challenges in 3D Reconstruction and Localization: Using the Resulting Geometry for Material Recognition

Presenter

  • Joseph M. Degol, The University of Illinois, Urbana-Champaign
November 29, 2017 - 10:00am to 11:00am

Abstract 

To autonomously navigate built environments (e.g., construction sites, bridges) for progress monitoring, a robot must be able to map the environment (3D reconstruction), pinpoint its location in that map (localization), and use the information to make decisions. In this talk, Joe will describe his work in overcoming challenges in 3D reconstruction and localization and using the resulting geometry for material recognition. He will focus on two topics: (1) ChromaTag, a colored fiducial marker and fast detection algorithm that enables detection at frame rates above 700 frames per second and (2) Exploring how sparse geometry can be used with appearance features for improved material recognition. He will also briefly touch on other work related to drones monitoring construction sites and grasp selection for prosthetic hands.

Additional Information 

About the Speaker:

 Joseph DeGol is a computer science PhD student at the University of Illinois. His interests in using geometry alongside images or video for reasoning about objects and camera motion in a scene have led to work in 3D reconstruction, visual odometry and camera pose estimation, and geometry-informed material recognition. He was awarded the National Defense Science and Engineering Graduate Fellowship, a 3M fellowship, and an NSF GRFP honorable mention. He also helped found Reconstruct, a startup that uses images of construction sites to map the progress on the site over time.

Sponsoring Organization 

Electrical and Electronics Systems Research Division

Location

  • Conference Center
  • Building: 5200
  • Room: Emory Room (214)

Contact Information

Share