Cluttered Scene Segmentation Using the Symmetry Constraint
PhD candidate, Computer Science
Advisors: Prof. Yiannis Aloimonos and Dr. Cornelia Fermüller
Rapid advances in robotic technology are bringing robots out of the controlled environments of assembly lines into the unstructured and unpredictable “real-world” workspaces of human beings. One of the prerequisites for operating in such environments is the ability to segment previously unseen objects. In this talk I will present an approach for object segmentation in 3D pointclouds of cluttered scenes based on the observation that the three-dimensional shape of common objects is bilaterally symmetric. Our approach consists of two tightly coupled steps. In the first step, candidate 3D bilateral symmetries are detected by extracting and matching surface normal edge curves of the pointcloud. In the second step objects are segmented by finding scene points that are consistent with a given symmetry candidate and at the same time satisfy the grouping principles of proximity and convexity. I will show the results of the experimental evaluation of this approach on a novel dataset of cluttered tabletop scenes and discuss why symmetry enables it to outperform current state-of-the-art approaches.
About the Robotics Graduate Student Seminars
The Robotics Graduate Student Seminars at the University of Maryland College Park are a student-run series of talks given by current graduate students.
The purpose of these talks is to:
- Encourage interaction between Robotics students from different subfields;
- Provide an opportunity for Robotics students to be aware of and possibly get involved in the research their peers are conducting;
- Provide an opportunity for Robotics students to receive feedback on their current research;
- Provide speaking opportunities for Robotics students.