A study of 3D object detection and classification

Author/Creator

Author/Creator ORCID

Date

2016-09-30

Department

Towson University. Department of Computer and Information Sciences

Program

Citation of Original Publication

Rights

Subjects

Abstract

This work intends to provide a vision and artificially intelligent recognition system design that can be used in a robotics platform for use in the virtual robotics competition as well as any platform requiring advanced visual processing. This work explores the DARPA robotics competition and analyzes the requirements of each task in the competition. These requirements set a good basis for developing requirements for visual object recognition system. Next we research a number of technologies that would be targeted for integrating into this system such as the Atlas sensor suite, the robotics operating system, Gazebo simulator, and the point cloud library. The system proposed expands on an existing system proposed by Bedkowski et al. and adds the ability to recognize more intricate, non-flat objects through the use of the point cloud library and robotics operating system.