Masters Theses

Date of Award

12-1996

Degree Type

Thesis

Degree Name

Master of Science

Major

Electrical Engineering

Major Professor

M. A. Abidi

Committee Members

Dan Koch, William R. Hamel

Abstract

A range image presents direct information of the location of sampled object points in three dimensions. In order for the range information to be useful, it must be registered with an existing model according to the position from which the data was acquired and the orientation of the sensor. Thus, given a finite number of range images in which the majority of an object's surface has been sampled, a prototype three-dimensional model of the object may be reconstructed.

The primary focus of the research described in this thesis has been to design and implement a system capable of automatically reconstructing a prototype three-dimensional model from a minimum number of range images of an object. Given an ideal 3-D model, our system iteratively renders range and intensity images of the object from a specified position, assimilates the range information into a prototype model, and determines a sensor position and orientation from which an optimum amount of previously unrecorded information may be acquired. The system tests at each iteration whether the model meets a given threshold of accuracy, at which point reconstruction is terminated. Such a system may find application in the context of robot navigation, manufacturing, or hazardous materials handling.

We have tested the system successfully on several synthetic data streams, and found each set of results to be reasonably consistent with an intuitive human search. The number of views necessary to reconstruct an adequate 3-D prototype will always depend on the complexity of the object or scene and the initial data collected. The models which the system recovers compare well with the ideal models, given a reasonable number of views. Our results illustrate the fact that range information of physical objects can be employed to automatically reconstruct a satisfactory dynamic 3-D model at a minimal computational expense.

Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS