Autonomous robotic ground penetrating radar surveys of ice sheets; Using machine learning to identify hidden crevasses

This paper presents methods to continue development of a completely autonomous robotic system employing ground penetrating radar imaging of the glacier sub-surface. We use well established machine learning algorithms and appropriate un-biased processing, particularly those which are also suitable for real-time image analysis and detection. We tested and evaluated three processing schemes in conjunction with a Support Vector Machine (SVM) trained on 15 examples of Antarctic GPR imagery, collected by our robot and a Pisten Bully tractor in 2010 in the shear zone near McMurdo Station. Using a modified cross validation technique, we correctly classified all examples with a radial basis kernel SVM trained and evaluated on down-sampled and texture-mapped GPR images of crevasses, compared to 60% classification rate using raw data. We also test the most successful processing scheme on a larger dataset, comprised of 94 GPR images of crevasse crossings recorded in the same deployment. Our experiments demonstrate the promise and reliability of real-time object detection and classification with robotic GPR imaging surveys.