Automated DTM generation for low textured images using multiple primitive multiple-image matching (MPM)

Abstract

In the frame of our long term, interdisciplinary Nasca/Palpa project, conducted in a close cooperation between our group and archaeologists from the German Archaeological Institute (DAI, KAAK Bonn), our aim is to develop a GIS database well-suited for management and analysis of the archaeological and spatial data acquired during various field campaigns by the project partners. The region around the two localities of Nasca and Palpa, about 400 km southeast of Lima (Peru), is well-known for its famous geoglyphs, ground drawings which were carved into the desert sand during the Nasca culture period (200 B.C. – 650 AD), the so-called “Nasca lines”. In order to map the geoglyphs and derive preferably accurate Digital Terrain Models (DTM) and orthoimages suited for documentation, digital preservation and GIS-based analyses, three blocks of aerial images were acquired: Sacramento (211 images, B/W and Color) and San Ignacio (168 images, B/W) around Palpa and the Nasca block (401 images, B/W) covering the Pampa de Nasca. The actual photogrammetric work focuses on the Nasca block, aiming for a widely automated processing concerning aerial triangulation and DTM generation. Previous investigations have shown that commercial photogrammetric software systems were not suited for an accurate DTM generation, mostly due to the xeric characteristic of the landscape which leads to low texture and partly poor image quality. In order to overcome the shortcomings of commercial software available today, our inhouse software SAT-PP (Satellite Imagery Precision Processing) was used for DTM generation. SAT-PP, contrary to most commercial systems, applies a simultaneous, multiple primitive multi-image matching approach for point measurement. Therefore, an increased accuracy and reliability of the matched points can be expected. A comparison of the DTMs generated for several stereo models versus the results obtained by commercial systems as well as to manual measurements performed on an analytical plotter shows the high accuracy potential of the SAT-PP approach