Mae ymagweddau sefydledig at archwilio archaeolegol yn pwysleisio rhagchwilio yn y maes, wedi’i gefnogi gan ddata wedi’i bell-synhwyro, ond gwaith sy’n gofyn am dipyn o adnoddau yw hwn. Trafodir gwaith rhagarweiniol ar Ynys Arran, a elwir ar lafar gwlad yn ‘Scotland in Miniature’ (yr Alban ar raddfa fach), i ddatblygu mapio graddfa genedlaethol, gan gynnwys cyflwyno darganfyddiadau newydd, agweddau ar ddata aml-raddedig a dyhead i ddatblygu llifoedd gwaith wedi’u symleiddio sy’n rhoi’r pwyslais ar ddata wedi’i bell-synhwyro a chanfod gwrthrychau awtomataidd. Mae hyn yn gofyn am ddatblygu llifoedd gwaith newydd, ond hefyd am ailystyried agweddau ar arfer sefydledig.
Mae’r cynnydd mawr mewn data wedi’i bell-synhwyro yn yr Alban yn cynnig cyfleoedd ac yn gosod her. Er bod map orthoffotograffig cyflawn o’r Alban ar gael ers rhai blynyddoedd bellach, mae sganio laser o’r awyr (ALS) yn ddarniog, er bod dyhead i’w helaethu neu ei gwblhau hyd yn oed. Mae hyn yn codi her ynghylch sut y bydd Historic Environment Scotland, fel corff arolygu cenedlaethol yr Alban, yn ymdrin â’r cynnydd cyflym mewn data wedi’i bell-synhwyro wrth ddiffinio ymagwedd at fapio cenedlaethol.
The increasing availability of remotely sensed data in Scotland presents opportunities and challenges. While complete coverage of orthophotos have been available for a few years now, availability of Aerial Laser Scanning (ALS) is patchy, though there is an aspiration to greater or even complete coverage. This raises a challenge to how Historic Environment Scotland, as the national body of survey for Scotland, deals with the rapidly increasing remote sensed data in defining an approach to national mapping.
Established approaches to archaeological prospection emphasise field reconnaissance, supported by remote sensed data, but this is a relatively resource intensive operation. Preliminary work on the Island of Arran, which is colloquially known as ‘Scotland in Miniature’, to develop national-scaled mapping is discussed, including presentation of new discoveries, aspects of multi-scaled data and an aspiration to develop streamlined workflows that foreground remote sensed data and automated object detection. This is requiring development of new workflows, but also the renegotiation of aspects of established practice.
73. 1977
RCAHMS
Easter Ross, Ross and Cromarty District,
Highland Region
Ordnance Survey
RCAHMS
North Kyle, Kyle and
Carrick, Strathclyde Region
Ordnance Survey
RCAHMS
Stirling District, Central Region
Ordnance Survey
RCAHMS
Aerial Survey updates
(Nationwide)
74. 1978
RCAHMS
Stirling District, Central Region
RCAHMS
Easter Ross, Ross and Cromarty District,
Highland Region
RCAHMS
Dumbarton District, Clydebank District,
Bearsden and Milngavie District, Strathclyde Region
RCAHMS
Lanarkshire
Ordnance Survey
Ordnance Survey
RCAHMS
Aerial Survey updates
(Nationwide)
Ordnance Survey
RCAHMS
Lunan Valley, Montrose Basin, Angus District,
Tayside Region
Publication
McWilliam
76. 1980
RCAHMS
Orkney
RCAHMS
North Kyle, Kyle and
Carrick, Strathclyde Region
RCAHMS
Argyll
Ordnance Survey
RCAHMS
Berwickshire District,
Borders Region
RCAHMS
Ewesdale and Lower Eskdale, Annandale and
Eskdale District, Dumfries and Galloway Region
RCAHMS
Aerial Survey updates
(Nationwide)
Publication
Coastal Survey
Ordnance Survey
77. 1981
RCAHMS
Argyll
RCAHMS
Orkney
RCAHMS
Ewesdale and Lower Eskdale, Annandale and
Eskdale District, Dumfries and Galloway Region
RCAHMS
North Kyle, Kyle and
Carrick, Strathclyde Region
Ordnance Survey
RCAHMS
Aerial Survey updates
(Nationwide)
78. 1982
RCAHMS
Argyll
RCAHMS
Orkney
RCAHMS
North Kyle, Kyle and
Carrick, Strathclyde Region
Ordnance Survey
Strathclyde
RCAHMS
South Kincardine, Kincardine and Deeside District,
Grampian Region
Ordnance Survey
Caithness
RCAHMS
Aerial Survey updates
(Nationwide)
79. 1983
RCAHMS
Argyll
RCAHMS
Orkney
RCAHMS
North Kincardine, Kincardine and Deeside District,
Grampian Region
RCAHMS
Angus District, Tayside Region
RCAHMS
Rhum, Lochaber District,
Highland Region
RCAHMS
North Kyle, Kyle and
Carrick, Strathclyde RegionRCAHMS
Aerial Survey updates
(Nationwide)
80. 1984
RCAHMS
West Rhins, Wigtown District,
Dumfries and Galloway Region
RCAHMS
Argyll
RCAHMS
Orkney
RCAHMS
North Kincardine, Kincardine and Deeside District,
Grampian Region
RCAHMS
Angus District, Tayside Region
RCAHMS
Aerial Survey updates
(Nationwide)
Publication
Hunter
81. 1985
RCAHMS
Argyll
RCAHMS
North Kyle, Kyle and
Carrick, Strathclyde Region
RCAHMS
West Rhins, Wigtown District,
Dumfries and Galloway Region
RCAHMS
Orkney
RCAHMS
Aerial Survey updates
(Nationwide)
Publication
Edinburgh
University
89. 1993
RCAHMS
Den of
Boddam
RCAHMS
Mar Lodge
RCAHMS
Upper Strathnairn
RCAHMS
Strathbraa
n
RCAHMS
OS
Revision
RCAHMS
Finlaggan
RCAHMS
East
Dumfries
RCAHMS
OS
Revision
RCAHMS
Aerial Survey updates
(Nationwide)
Archaeology Scotland
Discovery and Excavation Scotland
(Nationwide)
Publication
Hood
114. CONTEXT
• Much of Scotland’s archaeology is not on
record
• Majority of RS data is not exploited for
archaeological purposes
• RS data is proliferating
• Use of RS data is largely constrained within
traditional methods of prospection & object
identification developed in the 20th century
• Implications of digital/3D? Can we play a
lead role?
116. NATIONAL SCALE ARCHAEOLOGICAL SURVEY:
SOME ISSUES
• Is ‘national coverage’ necessary?
• Balance between extent of coverage and level of recording detail
• How comprehensive? (i.e. what sources are consulted?)
• What comprises a minimum level of record?
• How far can RS data can be relied on as the primary source?
• What is the role of field observation?
• How do we define ‘quality’?
• How can we ‘speed up’ archaeological survey?
119. LEVELS OF SURVEY
Where we are at the moment...
• Level 0 - Preliminary, desk-based survey and record
• Level 1 - Basic record: Field visit to produce an accurate location
at basic mapping scale, brief description and classification
• Level 2 - Landscape mapping: detailed landscape mapping by
EDM/GPS 1:2500, 1:5000; 1:10,000) allowing features to be
seen in context
• Level 3 - Groups of monuments: group of sites or structures
planned by SRA at 1:250 and 1:500
• Level 4 - Individual detailed survey and record using plane table
Where does Arran survey and proposed
national mapping fit in?
120. DO BASIC RECORDS FIT MULTIPLE PURPOSES?
• Where, what, when and how – when do we need more?
• Relationships with ‘deeper’ (sometimes) records in HERs
• How often do users find information exactly the way they want it?
• Is NRHE ‘best’ as an index?
• Would readily available ALS visualisations change what we do?
124. ARE THERE ISSUESWITH A PRIMARY RELIANCE
ON RS DATA FOR THE NATIONAL SURVEY?
• Are we happy to use RS data alone for site creation/upgrade?
• How do we develop our confidence in using ALS data?
• How do we assess RS data reliability?
• Do we believe in ‘ground truth’?
• How do we use ground observation selectively?
• What is the role(s) of field visits?
125. NR83SE 26 8908 3026
NR 8908 3026. A heather-covered hut circle was
found during field investigation situated on gently
sloping moorland at approximately 300ft OD. The
earth and stone bank, spread to 3.0m and up to
0.8m high, encloses a level interior 5.5m in diameter.
There is a 1.0m wide entrance in the south-east with
a single orthostat on its east side.
Approximately 100.0m south of the hut circle is an
80.0m long bank. Comprising earth and large stones,
it is 2.0m wide and 0.4m high and is probably a field
bank associated with the hut.
Lying between the hut circle and the bank is a group
of 15 heather-covered stony mounds similar to other
probable clearance heaps in this area (see NR83SE
11). Visited by OS (TRG), 20 October 1977
126. A heather-covered probable hut circle was found during
field investigation situated on a gentle slope in open
moorland. The earth and stone bank, spread to 2.5m and
up to 0.8m high, encloses a level interior 5.0m in
diameter. In the south-west the bank is 5.0m wide and
1.8m high, possibly the result of dumping from nearby
peat cutting. There is a 1.5m wide entrance in the south-
east; a small gap in the north-west appears to be a
modern mutilation.
Visited by OS (TRG) 20 October 1977
127.
128.
129. A basic record
Where
NR 92193 38348
What
Shieling hut(s)
Shieling mound(s)
When
Post Medieval
Event
‘An archaeological survey
of Arran was conducted
between ** and ** by HES,
Survey and Recording ....
ALS data ..., extent and
Level of Survey...’
130. A more detailed record
Where
NR 92193 38348
What
Shieling hut(s)
Shieling mound(s)
When
Post Medieval
Event
‘An archaeological survey of Arran was conducted between ** and ** by HES,
Survey and Recording .... ALS data ..., extent and Level of Survey...’
A scatter of at least 25 shieling mounds have been identified from Airborne
Laser Scanning data derived visualisations beside the Allt Airigh na Cuiseig in
Glen Iorsa. The majority of the mounds lie to the S of the stream, and are
formed as distinct grassy mounds, with a depression, sometimes
subrectangular, or dimple on the summit. The largest of the mounds measures
about 6m in diameter and up to 0.8m in height.
Information from HES Survey and Recording (DCC) 24 October 2017
131. WHAT ARRAN SURVEY PROJECT LOOKS LIKE?
Maximise use of RS data
• Prospection/reconnaissance
• Level of confidence
• Site/record creation to minimum standard
Where (coordinate/polygon)
What
When
Event
• Identify sites/areas for field visits (anomalies, gaps, poor ALS data)
Strategic field work
• Visit anomalies
• Targeted field reconnaissance in ‘gaps’
• Some more intensive RS-based mapping and field investigation
Develop digital documentation/interpretation experience/skills
132. SOURCES:
Old and current OS maps
First Edition Settlement Project
Aerial Photographs
Historic Land Use Assessment
(HLA)
ALS visualisations
FROM POINTTO POLYGON
136. TAKING THIS AS A MOMENT TO REFLECT
ON THE KNOWLEDGE PRODUCTION
PROCESS
• Role of expertise
• How we think about observation
• Black boxes and explanation
• Descriptions of reliability / probability
• When are we data driven and pattern seeking?
When are we fitting to conceptual models?
137. TO BETTER DEFINETHE ESSENCE OF
WHAT COMPRISESTHE
ARCHAEOLOGICAL IN LANDSCAPE
SURVEY
What we see is dependent on:
• our expectations
• our experience
• our questions
• our acuity
139. Give half a dozen people the same
information and you get half a dozen different
interpretations, all equally valid and probably
not even mutually contradictory in any
significant way, but each of them will bear the
style of the individual interpreter
Reginald Hill, 2008.The Roar of the Butterflies
143. RAPID DEVELOPMENT ACROSS MANY FIELDS
Medical imaging
Cell detection &
Classification
3D
Histopathology
Surveillance
Automated
Video
Surveillance
Systems
Target detection
Social media
Face recognition
Automated
tagging
How variable and
complex are
archaeological
remains compared to:
• Puppies
• Human Faces
• Medical images
144. Trier (2016) and Zingman (2016)
• Pre-training on ImageNet
• Then train (learn) on a relatively small target set of images
• Can use Matlab toolbox Mat-ConvNet
Automatic charcoal kiln detection for a 3×3 km test
area.
Convolutional Neural Networks (Machine Learning)
Underpinned by field observation
Image courtesy of Ø. Trier
145. AUTOMATED OBJECT DETECTION: PROOF OF
CONCEPT
• Norwegian Computing Center
• Height data
• Learning sets for hut-circles, small cairns and shieling huts
• Development for national mapping, crucial for big data
• Contribute to understanding better how we see sites
• No threat to archaeological interpretation & expertise, but a tool
146. AUTOMATION ON ARRAN
Manual mapping
• Up to 30km2 per person/day for basic
polygonisation only
• Single data source
CNN on Arran
• Learning sets (hut-circles, small cairns,
shieling huts)
• Work in progress
148. Procedures for rapid assessment of ALS/RS data as it becomes
available within a strategy for national mapping
RS data at the forefront, multiple scales of survey
Methods to express varying confidence in outputs from RS
Review what constitutes an adequate archaeological record
Assessment of the potential representativity of data
Creation of a ‘learning set’ for automated object detection
routines for ALS data
Developing an approach to national mapping
149. ACKNOWLEDGEMENTS
• ALS data: Fugro and Scottish Government
• Adara López-López
• Žiga Kokalj
Thank you for your attention!