On Choosing Training and Testing Data for Supervised Algorithms in Ground Penetrating Radar Data for Buried Threat Detection

摘要

Ground penetrating radar (GPR) is one of the most popular and successfulsensing modalities that has been investigated for landmine and subsurfacethreat detection. Many of the detection algorithms applied to this task aresupervised and therefore require labeled examples of target and non-target datafor training. Training data most often consists of 2-dimensional images (orpatches) of GPR data, from which features are extracted, and provided to theclassifier during training and testing. Identifying desirable training andtesting locations to extract patches, which we term "keypoints", is wellestablished in the literature. In contrast however, a large variety ofstrategies have been proposed regarding keypoint utilization (e.g., how many ofthe identified keypoints should be used at targets, or non-target, locations).Given the variety keypoint utilization strategies that are available, it isvery unclear (i) which strategies are best, or (ii) whether the choice ofstrategy has a large impact on classifier performance. We address thesequestions by presenting a taxonomy of existing utilization strategies, and thenevaluating their effectiveness on a large dataset using many differentclassifiers and features. We analyze the results and propose a new strategy,called PatchSelect, which outperforms other strategies across all experiments.