Using improved outlier estimation for hyperspectral target detection Conference Paper uri icon

abstract

  • We present a thorough examination of different noise estimation methods for usage with target detection algorithms for hyperspectral datasets. The different algorithms were designed with two approaches for dealing with outliers: The first allows outliers to contribute beyond their actual population to the estimated distribution, while the second approach limits them. In Addition, the matched filter distribution on the eigen-direction was analyzed using PCA for each algorithm, presenting a novel way to compare and examine the behavior of target detection method.

publication date

  • January 1, 2017