Automatic Near Surface Estimation from Radar Imagery

The Center for Remote Sensing of Ice Sheets has developed radars for NASA's Operation Ice Bridge program in order to map near-surface internal layers for estimating the accumulation rate. In snow and ice, internal layers are created by changes in the ambient conditions at the time of deposition, and represent contrasts in density, electrical conductivity, and ice crystal orientation. By identifying and tracing internal layers in radar images of the Antarctic snow cover, these layers can be used to measure snow accumulation over time. Scientists have manually traced layers in large data volumes, and it requires time-consuming sparse hand-selection and interpolating between selections to save time. An automated algorithm will allow for studying more images and developing models to reconstruct and forecast ice sheet dynamics. We have developed an approach for automatically estimating near surface layers in snow radar echograms using a computer vision technique. The approach uses active contour models, which finds high-intensity edges likely to correspond to layer boundaries, while simultaneously imposing constraints on smoothness of layer depth and parallelism between layers. Results are evaluated and presented using metrics of accuracy and computation time.