School of Computing Faculty of Engineering

Seeing the unseen:

Improving aerial

prospection outside

the visible spectrum

David Stott, Anthony Beck,

Doreen Boyd & Anthony Cohn

Overview

• An introduction to the DART project

• The problem

• Contrast

• Principles of detection

• Preliminary results

• Lots of graphs

• Further work

• Problems

• Proposed analyses

The DART project

• Detecting Archaeological Residues using remote Sensing

Techniques

• Soil properties

• University of Birmingham

• University of Winchester

• Geophysics

• Bradford University

• Optical (aerial and satellite detection)

• University of Leeds

• University of Nottingham

How do we detect archaeological features?

•Contrast with the background

•Changeable:

• Land use

• Cultivation regime

• Vegetation

• Species & variety

• Growth stage (phenology)

• Soils

• Weather

The problem

• Observer directed aerial photography

• Bias

• Soils (Jessica Mills & Rog Palmer)

• Honeypots (Dave Cowely & Kenny Brophy)

• Visible spectrum

• Sensors

• Underutilised because we don’t know how best to use them

• Hyperspectral

• Focus on data reduction

• Very few archaeologically commissioned flights

• Thermal

Some rights reserved by ZakVTA

Aims

• To understand how archaeological features interact with

and influence the surrounding environment

• If we do this we can work out how to detect them better

• Improved exploitation of existing sensors

• Improved development of new sensors

This aims of this project are:

• To identify optimal timing for acquiring aerial and satellite

imagery for archaeological prospection

• Commissioning new imagery

• Evaluating existing archives

What I’m doing: Fieldwork

• Measurements taken on transects across linear features on

at least a monthly basis

• Spectro-radiometry (more on this in a minute)

• Surface properties

• Vegetation coverage (near vertical close-range photography)

• Vegetation growth stage

• Height

• Feekes scale

• Vegetation density

• Leaf Area Index (LAI)

Spectra-ma-what-now?

• Spectroradiometry

• ASD FieldSpec Pro

• Produces a spectral profile

• 350nm-2500nm (Visible-Short Wave Infrared)

• c. 1.4-2nm Sampling interval interpolated to 1nm

• Usable 2hrs either side of solar noon

• Needs clear-ish skies

Flights

• Environment agency

• CASI

• High spatial resolution ortho-photography

• 28th June 2011

• NERC ARSF

• Eagle (visible – near-IR) & Hawk (near-IR – SWIR)

• High spatial resolution ortho-photography

• Thermal?

• 14th June 2011, 23rd March 2012

• 3 further flights during 2012

Problems…...

Problems

• 2011 Driest spring in eastern England for 100 years

• Extreme conditions

• 2012 due to be an even more extreme drought

• I want it to be a bad spring and a worse summer (sorry. Kind of)

• Not much subtlety in the vegetation marks…

“Why do you need hyperspectral when you can see the cropmarks on the ground like this”

Solution: Extend temporal depth?

• Can I use lower spatial resolution satellite data?

• Paleochannels as a proxy for archaeological vegetation marks?

• Need to test this

Further work: Analysis

• Building an ontology

• Identifying diagnostic absorption features

• Well known from precision agriculture & remote sensing

• Using this to evaluate contrast

• Python code to compare spectra

• Field spectra (high temporal resolution, low spatial coverage)

• Aerial spectra (low temporal resolution, high spatial coverage)

• Correlating contrast to environmental variables

• Weather

• Soil moisture

Further work: Building a knowledge-based system

• Testing

• Using this to predict contrast in 2013

• Using this to predict contrast in archive imagery

• NERC flights?

• Geoeye satellite data?

• Aerial photos?

Finally

• DART is Open Science!

• PLEASE re-use our data

• Servers online spring-summer 2012

• www.dartproject.info

• @DART_Project

• http://www.flickr.com/groups/dartproject/