<title>Scene/object classification using multispectral data fusion algorithms</title>

Abstract

Near-simultaneous, multispectral, coregistered imagery of ground target and background signatures were collected over a full diurnal cycle in visible, infrared, and ultraviolet spectrally filtered wavebands using Battelle's portable sensor suite. The imagery data were processed using classical statistical algorithms, artificial neural networks and data clustering techniques to classify objects in the imaged scenes. Imagery collected at different times throughout the day were employed to verify algorithm robustness with respect to temporal variations of spectral signatures. In addition, several multispectral sensor fusion medical imaging applications were explored including imaging of subcutaneous vasculature, retinal angiography, and endoscopic cholecystectomy. Work is also being performed to advance the state of the art using differential absorption lidar as an active remote sensing technique for spectrally detecting, identifying, and tracking hazardous emissions. These investigations support a wide variety of multispectral signature discrimination applications including the concepts of automated target search, landing zone detection, enhanced medical imaging, and chemical/biological agent tracking.