Two methods for semi-automated feature extraction from lidar-derived DEM designed for cairn-fields and burial mounds

Benjamin Stular

Abstract


We are in agreement with the session call that among others a reason hindering more efficient emergence of semi-automated or supervised detection techniques to identify anthropogenic features on remote-sensing data are critics stressing the irreplaceability of human judgement in recognising archaeological features. In the case of the lidar-derived data, it seems, the prevailing reason is the fact that archaeological features come in a near-unlimited assortment of shapes and sizes, though. Thus, the successful efforts so far have been focused on a limited number of homogenous feature types that appear in great quantity, ie. roads, open mining shafts or cairn-fields.
We are presenting two methods developed for semi-automated detection of individual cairns within a cairn-field. The first method is based on the standard-deviation-of-elevation based local relief and subsequent classification of 2D shapes. The second method is based on peak finding algorithm. Both methods are implemented in existing free GIS software packages.
The pipeline for the two methods will be presented. The results will be showcased and discussed on two different case studies aiming at providing not just an "ideal" conditions but also a very demanding one.