Date of Award

Spring 5-30-2024

Document Type

Thesis (Undergraduate)

Department

Computer Science

First Advisor

Xenia Dela Cueva

Abstract

This paper presents a method for processing data for a mechanical scanning sonar to use with a monocular camera. As the field of underwater robotics expands, it becomes increasingly important to develop techniques for determining the robot's pose using cost-effective sensors such as sonars and cameras. Traditional visual odometry and visual SLAM methods face limitations in underwater environments due to unique challenges posed by light interaction with water. Therefore, integrating additional sensors like the mechanical scanning sonar is essential. This research focuses on determining the extrinsic transformation between the sonar and the monocular camera. By leveraging existing knowledge of sensor placements and camera-to-image models, sonar readings are projected onto underwater images. The projected data is then corrected, and an effort was made to evaluate depth estimation for features used in camera pose estimation between images in subsequent timesteps.

Available for download on Friday, May 30, 2025

Share

COinS