3D mapping with affordable low altitude devices: a case-study in N Portugal

This case is study aimed at evaluating the applicability for 3D documentation of archaeological features of a “Hellkite” equipped with a compact digital camera, combined with software based on “structure from motion” (SfM) processes. The work shown in the case study was developed in the Iron Age hillfort of São Vicente da Chã (Trás–os-Montes, Northern Portugal). The aim of the project was the identification and 3D documentation of a series of structures in the surroundings of the hillfort that are thought to be related with ancient mining activities. The site is located within a dam, and is usually sunken. However, after a period of draught in the summer-autumn of 2011, a series of structures in bare rock were exposed that were in exceptional condition to be documented. 

Case study: the site

This work is part of a doctoral research (João Fonte), aimed at the analysis of the archaeological landscapes of the Late Iron Age and Early Roman period in Northwestern Iberia.

Among the settlement sites of the study area, the hillfort of São Vicente da Chã makes a singular case. It is located in the hilly province of Trás-os-Montes in Northern Portugal.

Nowadays the site falls within the Alto do Rabagão dam, completed in 1964, whose reservoir covers an area of about 2200 hectares and has a maximum volume of 569 hm3. Most of the site environs are usually sunken.

 A first piece of work was the reconstruction of the original, pre-dam topography, through the photogrammetric restitution of aerial images from 1949. A 5m DEM and a 50 cm orthomosaic were obtained [1].

Archaeological problems: objectives

The target: being usually sunken, no soil or vegetation exist around the site. A decrease in the water level in the summer-autumn of 2011 revealed a number of features (pits, trenches, holes, etc) cut on the bedrock and exceptionally visible.

Features of this type are but rarely know in this area and period. They are supposed to be connected with mineral (possibly tin) extraction and/or processing.

The aims of the work reported here were:

to record all those exceptional features for their further analysis. Documentation should be quick, reliable and affordable;
to contrast the results of different field methods: GPS vs. digital aerial photos and “structure from motion” (SfM) processing [2, 3].

How it was made: methodology

Two kind of data were captured in the field, following a process well described elsewhere [3]:

• Low-altitude aerial photos from a Helikite aerial photography system. The camera was a Ricoh GR Digital II, 10 MP, 28 mm GR wide-angle lens, CCD sensor.
• GPS measurements, both for detailed mapping of discrete structures (pits, holes,...) and for GCP prupose. Two Leica 1200 were used in a dynamic base-rover configuration, with RTK correction. Measurement accuracy was >1cm.

Data capture involved 3 people for 8 hours, data processing took another 8 hours.

Software used for GPS data processing, photo processing and accuracy assessment were respectively:

• Leica GeoOffice
• Agisoft Photoscan Professional 0.8.5
• ArcGIS 10.

What we got: results

An area of ca. 470 sq.m. was documented through 63 aerial photos. The SfM process produced a total of 215.412 points from which a 0,0174762 m. DEM and a 0,0105584 m. orthoimage were derived.

 Accuracy of the DEM was measured against a sample of 755 GPS measurements, most of them around and inside the archaeological features.

Results of error measurement are (in m.): 

This is well in line with what was obtained in similar test before [4]. Higher errors were concentrated, as expected, along breaklines, but also in the darker areas (shadows) on the photos.

A simple procedure to check that likely correlation was developed:

• An IHS image (Intensity, Hue, Saturation) was obtained after the RGB orthophoto
• Component 1 (Intensity) was compared to the DEM error for all the 755 control points.

Final remarks

SfM proves once again to be both a highly reliable and affordable methodology for archaeological documentation.
In-between the creation of large-scale topographic surfaces or small-scale 3D models of objects, it is also well suited for the documentation of archaeological features in survey or excavation contexts (as already tested, [4]).
Issues with lighting/shadowing in photos must be considered to avoid loss of accuracy or voids in the final results. Additional photos taken at ground level could solve that issue.

References

[1] Fonte, J., P. Redweik, J.A. Gonçalves, L. Lobos, and M. Felipe. 2012. Recovering Missing Landscapes Through Historical Aerial Images. In EARSeL Workshop 2012 "Advances in Remote Sensing for Archaeology and Cultural Heritage Management". Ghent (Belgium).

[2] Verhoeven, G., M. Doneus, Ch Briese, and F. Vermeulen. 2012. Mapping by matching: a computer vision-based approach to fast and accurate georeferencing of archaeological aerial photographs. Journal of Archaeological Science 39: 2060-2070.

[3] Verhoeven, G., D. Taelman, and F. Vermeulen. 2012. Computer Vision-Based Orthophoto Mapping Of Complex Archaeological Sites: The Ancient Quarry Of Pitaranha (Portugal–Spain). Archaeometry (available on-line first; doi: 10.1111/j.1475-4754.2012.00667.x)

[4] Doneus, M., G. Verhoeven, M. Fera, Ch. Briese, M. Kucera, and W. Neubauer. 2011. From deposit to point cloud - A study of low-cost computer vision approaches for the straightforward documentation of archaeological excavations Geoinformatics (XXIIIrd International CIPA Symposium) 6: 8–188.

Source

João Fonte, Pastor Fábrega-Álvarez, César Parcero-Oubiña and Alejandro Güimil-Fariüa. 3D mapping with affordable low altitude devices. A case-study on the documentation of archaeological features in the surroundings of an Iron Age hillfort in Northern Portugal. Poster presented at AARG - Aerial Archaeology Research Group Conference, 2012 (Budapest, Hungary).