Reconstruction of snow and ice surfaces using multiple view vision techniques

In this paper we describe a vision system for reconstructing the relative structure of snow and ice surfaces. This effort is one step toward the long-term goal of rendering absolute scale topography. The target application is mapping of sea ice topography from transiting ships for monitoring sea ice mass balance. Images of snow and ice scenes pose difficulties in image analysis due to low contrast and the presence of large areas with low color variation. Since this is a less studied area in computer vision, our first goal is to obtain a candidate data set, evaluate different camera configurations, and develop algorithms to process the data. For simplicity, we chose the imaging of river ice as a surrogate for sea ice because of its accessibility. Multiple views of a scene were obtained using a variable zoom camera (15-300mm) mounted on a pan-tilt unit which moved along a 1m baseline. Data were collected at seven locations from heights of 2-100m and viewing angles of...

• Type: Conference
• Year: 2008
• Where: EASTERN SNOW CONFERENCE
• Authors: GOWRI SOMANATH , ROHITH MV , CATHLEEN A. GEIGER , AND CHANDRA KAMBHAMETTU
• Keywords: EASTERN SNOW CONFERENCE 2008, Sea Ice Reconstruction, Stereo, Texture-less Stereo