Using Drones For Field Survey

Images taken by cameras mounted on small unmanned aircraft may now be photogrammetrically processed without requiring access to expensive software and specialised hardware. Such techniques have enormous potential for the visualisation of landscapes, sites and monuments and are being applied in a number of ways by Historic England’s Imaging and Visualisation Team. Here we outline one of these applications: the use of images captured by such aircraft to assist in the making  of archaeological field surveys.

This is done using a technique known as structure-from-motion (SfM). This matches images of objects taken from different overlapping viewpoints so as to create a digital model of a site or landscape.

The matching of images is based on pixel patterns in the images themselves and the calculated locations of the cameras which took them; it also takes into account distortions created by the lenses used.

Photogrammetric multi-view stereo algorithms are then used to place the pixels to form a digital model of three-dimentional space. The result is a digital surface model of the monument or landscape depicted in the original images. Everything from entire landscapes to small objects can be represented in this way.

Input imagery is not limited to that generated by normal cameras – near-infra-red and other imagery from beyond the visible spectrum can be processed, as can historical imagery, provided (amongst other things) that the images overlap with each other sufficiently.

A field survey plan produced in this way was recently tested by Historic England at Thornton Abbey in north Lincolnshire. Now chiefly known for of its large and ornate fortified monastic gatehouse, Thornton Abbey was the subject of research by English Heritage between 2007 and 2010 (Oswald et al 2010). One product of this was a detailed, and conventionally produced, archaeological field survey of the earthworks that lie between the former claustral buildings and the gatehouse. With a good recent field survey already in existence, this was an ideal site in which to assess the effectiveness and accuracy of the new technique. The area surveyed for the 2010 report measured approximately 500 metres × 250 metres ; it was flown by the small unmanned aircraft and the results compared with the traditional survey.

The earthworks being surveyed provide evidence for medieval buildings, for the landscaping associated with the creation of a stately home on the site by Sir Vincent Skinner in about 1607, and for a number of 19th- and 20th-century archaeological excavations. The entire precinct is a scheduled ancient monument, while the main part of the site, including the gatehouse and ruins of the church and claustral buildings, has been in state guardianship since 1938.

The site was flown by a subcontractor, Skyline Images Limited, who used a Droidworx Aeronavics XM8 octocopter, carrying a Canon EOS 5D Mk III digital SLR camera. The amount of ground surface represented by each pixel (ground sample distance) was specified at 40 millimetres, which meant flying at the legal ceiling of 120 metres. The site was covered by 56 vertical or near-vertical shots. The brief specified a front-to-back overlap between images of at least 80 per cent and a side-to-side overlap of at least 60 per cent. A small number of oblique images were also taken.

Immediately prior to taking the images, a network of ground control points was established across the site. These are used to optimise alignment of the aerial images and to place the survey accurately 'on the map'. As paint marks could not be used (livestock were present on the site, and the abbey's summer opening to the public was imminent), paper plates were pinned to the ground using survey pegs. These proved very effective: they are cheap, clearly visible and unambiguous in the photography itself. They are also easy to place and remove.

The imagery was processed using Agisoft PhotoScan Pro software. After SfM alignment and filtering a sparse point cloud of approximately 250,000 tie points was produced. Next, a denser point cloud was generated and control points added.

This point cloud was then classified, so that features such as trees, scrub and buildings could be separated from the ground surface itself. The result was a digital elevation model (DEM) which was exported for analysis in a GIS; and a composite orthoimage – an image in which the scale is uniform – which was derived from all the input images.

This DEM was then analysed using ArcGIS software so as to create a composite hillshade note that the techniques employed are straightforward and could be replicated in many open-source alternatives. This elucidated many of the ground variations hinted at in the raw DEM. A slope analysis was also made: indeed, many other analytical processes could have been applied to the data.

As hoped, both the DEM and the hillshade image correlated very strongly with the hachured plan produced by the 2010 survey. It seems, then, that the technique has potential use as an aid to the survey and investigation of archaeological landscapes. Indeed, the analyses highlighted certain features and areas not previously noticed, and which may benefit from further examination.

These results were encouraging and it was decided to test the process further at a site which had not been surveyed before. This was the scheduled former lead mine at Ashnott in Lancashire, then a Heritage at Risk site on a small limestone knoll on the edge of the Hodder valley north of Clitheroe, in the southern part of the Forest of Bowland AONB. Historic England's Assessment Team had been asked to survey the site as part of plans to improve its future management.

Documentary research suggests that Ashnott mine could have been active around 1300; it was certainly a going concern when Thomas Proctor entered into a three-year lease with the Duchy of Lancaster to 'digge, take & myne leade' at 'Asshe Notte' in 1538 (The National Archive 1538-9). By the time the mine closed in the 1830s, the victim of a general slump in lead prices, generations of miners had left behind a tightly-knit complex of surface workings and underground levels by chasing the erratic patterns of mineralisation throughout the knoll.

The survey aimed to understand how this mine developed, and to ensure that new fences, intended to safeguard the remains by improving stock management, were correctly placed. It was also designed to highlight areas where the collapse of old, poorly-sealed shafts presented a danger to livestock and hill-walkers.

The site was flown in a similar fashion to Thornton Abbey, this time by Aerovision UK using a fixed-wing SenseFly eBee SUA carrying a Canon Ixus point-and-shoot camera. The imagery was gathered and processed in a comparable way to that for Thornton Abbey, but in this case the digital models were used in AutoCAD to draft an outline plan of the earthworks similar to those employed in traditional earthwork surveys. This plan was then taken back into the field, where it was verified, refined, and augmented by close observation and the judicious use of survey-grade global navigation satellite system equipment. The resulting earthwork plan, with slopes expressed as hachures, is somewhat less detailed than that which would normally have been produced by traditional ground-based survey. It is metrically accurate, however, and sufficiently nuanced to support archaeological analysis of the site.

Crucially, this plan was perfectly adequate for the purposes required of it. Measured against the scale of survey standards published byHistoric England, in which Level 2 records the general form of a monument and Level 3 captures its full complexity, this SfM-derived method might sit at 2.5 or perhaps a little higher. It also took less than half the time that would have been required had traditional methods been used to create a comparable plan.

A word of caution, however, to others encouraged to pursue this approach. Detailed and highly flexible three-dimensional imagery is a tremendous tool, but interpretations derived from it must still be informed by an experienced eye if they are to be robust. From the surveyor's perspective, the most valuable parts of the process are the site visit before the flight, which helps develop a good understanding prior to mapping the patterns observed from the air, and the detailed reassessment of the SfM-derived plan once it has been taken back on site. Only then do the finer distinctions between such features as paths and watercourses, washing floors, and working areas become fully apparent.

Postscript

Since this article was originally published on 31 March 2015, the Ashnott site has been successfully taken off the Heritage at Risk Register. As noted in the Heritage at Risk report for 2017, the survey work was instrumental in this success.

Biographies

Further reading

Historic England 2017: Photogrammetric Applications for Cultural Heritage

English Heritage 2007 Understanding the Archaeology of Landscape: a Guide to Good Recording Practice

The National Archive, lease by King of Ashe Notte to Thomas Proctor, E210/1477 1538 -9

Oswald, A Goodall, J Payne, A and Sutcliffe, T 2010 Thornton Abbey, North Lincolnshire: Historical, Archaeological and Architectural Investigations. Swindon: Historic England Research Report Series 100-2010

Went, D 2014 Ashnott Lead Mine, Ribble Valley, Lancashire: an Archaeological Survey of the Landscape Evidence. Historic England Research Report Series 74-2014