Improving detection using a Material Pattern Matching technique in hyperspectral images

Hyperspectral sensing provides unique capabilities due its intrinsic ability to identify materials. However, like in any other pattern recognition application, system noise and variations in target property (spectral signature in this case), give rise to false alarms while processing. False alarms can also occur if the scene is cluttered with objects made of similar materials. In this study a simple method of Material Pattern Matching is described for improving the effectiveness of hyperspectral technology for the detection and recognition of rigid objects. Essentially this is a fusion of spatial and spectral information. An application of the method is demonstrated for a heavily cluttered maritime scene.