Date of Award:

5-2014

Document Type:

Thesis

Degree Name:

Master of Science (MS)

Department:

Electrical and Computer Engineering

Committee Chair(s)

Scott E. Budge

Committee

Scott E. Budge

Committee

Jacob Gunther

Committee

Don Cripps

Abstract

Three-dimensional (3D) imaging has found a lot of use in modern era, owing to the fast ongoing advancement in electronics, processors, and sensors, all available at affordable prices. Three-dimensional television, games, and 3D projectors are some of the commonly found utilities which use 3D imaging. 3D images used for surveillance, exploration, target acquisition equipment, etc., are few of the military applications of 3D imaging. As the need increases, more techniques are being introduced to create 3D images which improve upon previously stated techniques. This thesis proposes an algorithm which uses “texel” images, which are images that have digital imagery fused to 3D points, to stitch together a 3D image of an scene.

A brief overview of the methods of calibration developed previously is given, followed by a method to improve the quality of range information obtained from the lidar sensor, which is prone to errors common to the type of sensor used. These errors cause the range measurements to deviate from the true values.

The second half of the thesis briefly goes over the work previously done on 3D image matching to produce 3D images. A few changes are suggested in the method proposed previously. Additional improvement using epipolar geometry, which is a geometric interpretation of 3D projective geometry, helps improve the results obtained. A solution is suggested that makes use of the transformation computed initially to delete bad or incorrect points from the set of points which are used to compute the initial transformation. A solution obtained from this new set of correct points is more accurate than the one computed initially. This solution is used to generate more points in the point set, which are used to compute the final transformation.

The results show that the methods implemented previously and the ones suggested in this thesis provide accurate 3D images, at low computing cost.

Checksum

ee103d56189d9e4e325b2f5b88d72d7f

Share

COinS