1. Robotic Image Capture
Darin T. Rajan
rajand@purdue.edu
Dr. Daniel G. Aliaga
Department of Computer Sciences
I. ABSTRACT
Computer simulation of real-world environments is one of the great challenges of
computer graphics. The ultimate goal is to easily acquire a digital model of a space, that
later can be used for a variety of applications. In order to capture such a large
environment in a timely fashion, the method must be mostly automatic and, preferably,
able to be operated by an untrained technician, such as a caretaker of the site. Thus, the
research challenge is to develop the algorithms and systems necessary to capture and
manipulate computer models of large and complex real-world scenes. To do this we have
assembled an autonomous mobile platform that moves through an environment and
acquires images from viewpoints spaced every few inches. These images will be used for
reconstructing novel views of the captured environments. The entire control interface was
developed using the mobile platform’s API in an OpenGL environment to create a
software system with built-in advanced robotic capabilities. The robot control system
includes visual displays of the mobile platform states and image capture progress with a
real-time monitoring program to interactively control the robot’s speed and navigation,
the ability to follow predetermined pathways approximately through the environment,
collision avoidance with nearby objects and people by using its sonar ring, and predictive
path planning to select the best paths to follow through the environment. To briefly
outline future work, the robot will be deployed to real-world locations, for example the
Purdue Union Ballroom, to capture environments. Through its high-resolution digital
camera it will be capturing millions of images and terabytes of data. While doing this it
will also be precisely computing its location through a laser positioning system we will
be interfacing.