Detection of Spatially Unresolved (Nominally Sub-Pixel) Submerged and Surface Targets Using Hyperspectral Data

Abstract

Due to the United States dependency on maritime travel, the proliferation of efficient and inexpensive naval mines poses a tremendous risk. Current MCM technologies have narrow FOVs, preventing timely, wide-area searches. These technologies require the operator to be in proximity to the targets, a dangerous scenario made worse when in denied territory. In an effort to mitigate these risks, the use of a high altitude hyperspectral sensor is proposed. The operational ability of a hyperspectral sensor to detect sub-pixel surface and submerged mines in non-littoral environments was evaluated using visual inspection and two common anomaly detectors Mixture Tuned Matched Filtering (MTMF) and Reed-Xiaoli (RX). Due to the unavailability of the DoDs Spectral Infrared Imaging Technology Testbed (SPIRITT), ProSpecTIR-VS3, a sensor similar spatially and spectrally to SPIRITT was flown over a range offshore California. This experiment included three surface and three submerged targets, each with a 0.8 meter diameter. Both 0.5 and 1 meter spatial resolution data were collected, allowing for both a resolved and unresolved analysis. While both anomaly detection techniques have their flaws, the success of this study is in proving the usefulness of hyperspectral data for sub-pixel mine detection.