Clutter and Anomaly Removal for Enhanced TargetDetection

摘要

In this paper we investigate the use of anomaly detection to identify pixels to beremoved prior to covariance computation. The resulting covariance matrix provides a bettermodel of the image background and is less likely to be tainted by target spectra. In our tests, thismethod results in robust improvement in target detection performance for quadratic detectionalgorithms.Tests are conducted using imagery and targets freely available online~*. The imagery wasacquired over Cooke City, Montana, a small town near Yellowstone Park, using the HyMapV/NIR/SWIR sensor with 126 spectral bands. There are three vehicle and four fabric targetslocated in the town and surrounding area.