Context extraction for local fusion for landmine detection with multi-sensor systems

Abstract

We present a local method for fusing the results of several landmine detectors using Ground Penetrating Radar (GPR) and Wideband Electro-Magnetic Induction (WEMI) sensors. The detectors considered include Edge Histogram Descriptor (EHD), Hidden Markov Models (HMM), and Spectral Correlation Feature (SCF) for the GPR sensor, and a feature-based classifier for the metal detector. The above detectors use different types of features and different classification methods. Our approach, called Context Extraction for Local Fusion with Feature Discrimination(CELF-FD), is a local approach that adapts the fusion method to different regions of the feature space. It is based on a novel objective function that combines context identification and multi-algorithm fusion criteria into a joint objective function. The context identification component thrives to partition the input feature space into clusters and identify the relevant features within each cluster. The fusion component thrives to learns the optimal fusion parameters within each cluster. Results on large and diverse GPR and WEMI data collections show that the proposed method can identify meaningful and coherent clusters and that these clusters require different fusion parameters. Our initial experiments have also indicated that CELF-FD outperforms the original CELF algorithm and all individual detectors.