Computer vision for polar sciences

• Main Author: Sorensen, Scott
• Format: Thesis
• Language: English
• Published: University of Delaware 2017
• Subjects: Applied sciences; Computer vision; Sea ice; Stereo; Arctic; ice covered waters
• Online Access: http://udspace.udel.edu/handle/19716/21748

As the Arctic becomes a place of commerce and industry, operating safely and ecologically in the region is growing in importance. Vessels traveling in ice-covered waters must constantly maintain awareness of conditions in the immediate area as well as large-scale regional ice and weather conditions to ensure the safety of the craft, its cargo and its crew. One of the key ways of doing this is by standardized visual observation. Many ice-going vessels are equipped with a variety of camera systems including thermal imagers and CCTV cameras. While these cameras are often used in support of the vessel and their operation, humans are kept in the loop. As commerce and exploration in the Arctic increases, better techniques are needed for extracting pertinent information, visualizing key data, and interaction. ☐ In this work I present a few imaging systems used in polar regions, and a series of techniques for extracting high level information from these systems. This work is aimed at assisting in decision-making for crafts, and people operating in and studying this environment. To this end, I have developed a 3D camera system for long term deployment aboard vessels in ice-covered waters. The Polar Sea Ice Topography REconstruction System, or PSITRES has been deployed on three research expeditions and collected terabytes of image data. Processing this data requires new techniques to make the problem tractable and to deal with the challenging nature of the data. In addition to PSITRES data, I present images collected from a variety of other imaging systems that were operated in parallel to PSITRES during its deployments, as well as remote sensing data. Kambhamettu, Chandra Ph.D. University of Delaware, Department of Computer and Information Sciences