CropMark -- A Matlab toolbox for archaeological imaging spectroscopy

Abstract

When using airborne imaging spectroscopy (AIS) data for archaeological research, at least two broad problems need to be solved (assuming that the georeferencing can be handled correctly). The first problem relates to the visualization of the huge amount of available data and the fact that imaging spectroscopy usually yields a large amount of redundant data. This makes data mining approaches necessary. The second problem relates to data quality. Indeed, as the upwelling electromagnetic radiation is recorded in small bands that are only about ten nanometres wide, the signal received by the sensor is quite low compared to sensor noise and possible atmospheric perturbations. In the same way, the necessary high spatial resolution (i.e. small ground sampling distance) further limits the useful signal stemming from the ground. For these reasons, radiometric filtering techniques are mandatory, as otherwise the noise component deteriorates the extracted information. A user-friendly Matlab-based toolbox, called CropMark, was developed to enable the extraction of information out of the recorded hyperspectral data cube. As the main application of this toolbox is the field of archaeology, the aim was to visualize the data highlighting possibly occurring crop or soil marks. Powerful filters based on the Whittaker smoother (currently not available within commercial image processing software) were implemented. The user can visualize the sequence of individual bands in an animated way, or look at (the first few) principle components. Shape information such as the red edge inflection point is derived from spectrally smoothed and oversampled signatures giving new insights into crop vigour/crop stress. Additionally, various standard and optimized hyperspectral vegetation indices were implemented. Areas can be highlighted having a similar spectral signature compared to a user-selected pixel or region of interest. The user can further test the usefulness of a large set of edge detection algorithms. Import and export functions are available regarding a number of standard image formats.

Michael Doneus
Michael Doneus
Key Researcher
Geert Julien Joanna Verhoeven
Geert Julien Joanna Verhoeven
Vice Director, Senior Researcher