Fast computation of digital terrain model anomalies based on LiDAR data for geoglyph detection in the Amazon

Abstract

ABSTRACT The detection of pre-Colombian geoglyphs, geometric structures outlined by trenches or walls, from airborne LiDAR data is usually made by visual observation of the variation in elevation and commonly using additional hillshading. Depending on the area covered by LiDAR to inspect and the variation in elevation, this method can be time consuming and inaccurate as it required to constantly adjust the contrasts of the elevation image or the parameter of the hillshading function, and the user can miss some important archaeological features. Here, we present a method to enhance the anomaly of the terrain without using focal operations to normalize the elevation of each pixel in relation to its neighbours. An example is given for two areas covered by LiDAR and containing geoglyphs under the forest cover in the Amazon and with a synthetic LiDAR footprint over a simulated dense forested area containing four geoglyphs with different shapes and height/depth characteristics. The normalization enables to remove the influence of the landscape mean elevation, to highlight the fine anomaly of the terrain. The produced equalized images enable a fast visual assessment of relief anomaly and of the presence of geoglyphs in large LiDAR datasets of hundreds or thousands of LAS files.