Detection of hidden surface crevasses on glaciers is a vital process involved in over-snow traverses for science and resupply missions in Polar regions. There are several areas warranting improvement in the current protocol for crevasse detection, which employs a human-operated ground penetrating radar (GPR) on a mid-weight tracked vehicle. In this fashion, a GPR scout team must plan an appropriate crevasse-free route by investigating paths across the glacier. This paper presents methods supporting a completely autonomous robotic system employing GPR probing of the glacier surface. We tested and evaluated three machine learning algorithms on post-processed Antarctic GPR data, collected by our robot and a Pisten Bully in 2009 and 2010 at McMurdo Station. We achieved 82% classification rate for a linear SVM, compared to 82% using logistic regression and 80% using a Bayes network for contrast. We also discuss independent versus sequential classification of GPR scans, and suggest improvements to or combinations of the most successful training models. Our experiment demonstrates the promise and reliability of real-time object detection with GPR.