Linear 3-D transformations of Landsat 5 TM satellite images for the enhancement of archaeological signatures during the phenological cycle of crops

Abstract

A linear 3-D transformation that can be used for the enhancement of crop marks related to buried archaeological features is developed and presented in this article. The methodology is based on three steps: (1) recalculation of the ground narrowband spectroradiometric measurements to the multispectral Landsat 5 Thematic Mapper (TM) sensor, based on the relative response filter of the sensor, (2) application of principal component analysis (PCA) transformation in order to determine the initial axes used for the orthogonal transformation, followed by (3) a 3-D rotation of the PCA axes. The linear coefficients of the transformation were retrieved and adjusted to different phenological stages of the crops. The transformation was successfully evaluated using both in situ measurements and Landsat 5 TM images in two different archaeological case studies. The proposed transformation tends to enhance archaeological signatures better than other established vegetation indices or algorithms, while the methodology can be expanded to any other multispectral satellite images using only the visible and very near-infrared part of the spectrum.