Landmine detection using FLGPVAR images

Landmine detection is a challenging problem remains to be solved. Forward-Looking Ground Penetrating Virtual Aperture Radar (FLGPVAR) is a popular method to detect landmines that are made of plastic or have little metal content. And landmine detection using FLGPVAR is actually an object recognition problem. This paper proposes to use the AdaBoost algorithm added with feature selection in the iterations. The motivation comes from the fact that the feature selection can decrease the training error and increase the margins of training data. And the training error is divided into the probability and the false alarm rate. Minimizing the false alarm rate with constant probability of detection can decrease the probability of missing in the testing data. Experiment results based on clutter lane data collected at a test site corroborate the effectiveness of the proposed classification algorithm to increase the accuracy for landmine detection.