Date post: | 14-Dec-2014 |
Category: |
Education |
Upload: | dart-project |
View: | 264 times |
Download: | 0 times |
School of ComputingFaculty of Engineering
Using multi-temporal benchmarking to determine optimal sensor deployment: advances from the DART project.
David Stott, Ant Beck and Doreen Boyd
Twitter: AntArch (in using the hashtag #EARSeL)
3rd EARSEL Workshop: Advances in remote sensing for archaeology and cultural heritage management
19th to 22nd September 2012
Presentation overview
• Detection summary
• Why do we need the DART project?
• Ground observation benchmarking at DART
• Data examples
• Multi-temporal spectroradiometry
• Conclusions
Archaeological ProspectionWhat is the basis for detection
We detect Contrast: • Between the expression of the remains
and the local 'background' value
Direct Contrast:• where a measurement, which exhibits a
detectable contrast with its surroundings, is taken directly from an archaeological residue.
Proxy Contrast:• where a measurement, which exhibits a
detectable contrast with its surroundings, is taken indirectly from an archaeological residue (for example from a crop mark).
Archaeological ProspectionWhat is the basis for detection
Micro-Topographic variations
Soil Marks• variation in mineralogy and
moisture properties
Differential Crop Marks• constraint on root depth and
moisture availability changing crop stress/vigour
Proxy Thaw Marks• Exploitation of different thermal
capacities of objects expressed in the visual component as thaw marks
Now you see meNow you dont
Archaeological ProspectionSummary
The sensor must have:• The spatial resolution to resolve the feature
• The spectral resolution to resolve the contrast
• The radiometric resolution to identify the change
• The temporal sensitivity to record the feature when the contrast is exhibited
The image must be captured at the right time:• Different features exhibit contrast characteristics at different times
What’s the problem? ‘Things’ are not well understood
Environmental processes
Sensor responses (particularly new sensors)
Constraining factors (soil, crops etc.)
Bias and spatial variability
Techniques are scaling!• Geophysics!
IMPACTS ON• Deployment
• Management
What do we do about this?
Go back to first principles:• Understand the phenomena
• Understand the sensor characteristics
• Understand the relationship between the sensor and the phenomena
• Understand the processes better
• Understand when to apply techniques
What do we want to achieve with this?
Increased understanding which could lead to:• Improved detection in marginal
conditions
• Increasing the windows of opportunity for detection
• Being able to detect a broader range of features
DART: Ground Observation Benchmarking
Try to understand the periodicity of change• Requires
• intensive ground observation
• at known sites (and their surroundings)
• In different environmental settings
• under different environmental conditions
DART: Ground Observation Benchmarking
Based upon an understanding of:• Nature of the archaeological residues
• Nature of archaeological material (physical and chemical structure)
• Nature of the surrounding material with which it contrasts
• How proxy material (crop) interacts with archaeology and surrounding matrix
DART: Ground Observation Benchmarking
Based upon an understanding of:• Sensor characteristics
• Spatial, spectral, radiometric and temporal
• How these can be applied to detect contrasts
• Environmental characteristics
• Complex natural and cultural variables that can change rapidly over time
DART: it’s all part of a process
DART: it’s all part of a process
DART: Sites
Location• Diddington, Cambridgeshire
• Harnhill, Gloucestershire
Both with• contrasting clay and 'well draining'
soils
• an identifiable archaeological repertoire
• under arable cultivation
Contrasting Macro environmental characteristics
Show sites here
DART: Probe Arrays
DART: Probe Arrays
As design
As built
DART: Probe Arrays
DART
ERT
B’ham TDR
Imco TDR
Spectro-radiometry transect
DitchRob Fry
DART
ERT
B’ham TDR
Imco TDR
Spectro-radiometry transect
DitchRob Fry
DART: Field Measurements
Spectro-radiometry• Soil
• Vegetation
• Up to every 2 weeks
Crop phenology• Height
• Growth (tillering)
Flash res 64• Including induced events
DART: Field Measurements
Resistivity
Weather station• Logging every half hour
DART: Field Measurements
Aerial data• Hyperspectral surveys
• CASI
• EAGLE
• HAWK
• LiDAR
• Traditional Aerial Photographs
• UAV
DART: Laboratory Measurements
Geotechnical analyses
Particle size
Sheer strength
etc.
Geochemical analyses
DART: Laboratory Measurements
Plant Biology• Rate of germination
(emergence)
• Growth analysis
• Number of Leaves
• Number of Tillers
• Stem length
• Total plant height
• Drought experiment
• Chlorophyll a fluorescence
• Soil and leaf water content
• Root studies
• Root length and density.
• Root – Shoot biomass ratio.
• Total plant biomass
• Biochemical analysis: Protein and chlorophyll analysis.
• Broad spectrum analysis of soil (Nutrient content) and C-N ratios of leaf.
DART: Data so far - Temperature
DART: Data so far –Temperature
DART: Data so far – Earth Resistance
DART: Data so far – Earth Resistance
Remote sensing
Spectro-radiometry: Methodology
• Recorded monthly
• Twice monthly at Diddington during the growing season
• Transects across linear features
• Taken in the field where weather conditions permit
• Surface coverage evaluated using near-vertical photography
• Vegetation properties recorded along transect
• Chlorophyll (SPAD)
• Height
Fiber-optic probe
Tired old laptop
(needs an LPT port…)
Reflectance panel
Instrument
(20Kg of back
pain)
To the visible
To the visible....... And beyond (08/06/2011)
But what about time? (14/06/2011)
Senescing (29/06/2011)
Senescant (15/07/2011)
Some rights reserved by ZakVTA
Analysis
• We are looking at relative contrast
• Identifying quantitative differences in the density of vegetation
• Identifying qualitative differences in vegetation stress & vigour:
• How to make this independent of density?
• Accounting for minor variations• Making sure things are comparable
• Illumination geometry
• Methodological blunders
Vegetation indices
• Mostly simple ratios
• Chlorophyll & biomass
• Carotenoid / chlorophyll• Photo-chemical Reflectance Index (PRI)
• Structure Insensitive Pigment Index (SIPI)
• Plant Senescance Reflectance Index (PRSI)
)570531(
)570531(
RR
RRPRI
Continuum removal
• Methodology explored by Kokaly & Clark (1999) and Curran et al (2001)
• Used to quantify leaf biochemical properties
• Uses diagnostic absorption features• Chlorophyll a+b
• Lignin
• Cellulose
• Proteins
• Water
Continuum removal
Designation(nm)
Start(nm)
Centre(nm)
End(nm)
Indicates
470 408 484 518 Chlorophyll a, b.
670 588 672 750 Red edge, stress
1200 1116 1190 1284 Water
1730 1634 1708 1786 Lignin
2100 2006 2188 2196 Nitrogen, starch
2300 2222 2306 2378 Nitrogen, protein, lignin
Continuum removal
• Band Normalised to depth of Centre of absorption feature (BNC)
• Band Normalised to Area of absorption feature (BNA)
))/1/())/(1( icci RRRRBNC
Conclusions
• Successful vegetation-mark detection depends on identifying the influence of the archaeological feature on its surroundings.
• Hyper-spectral remote sensing enables us to look for specific indications of this influence.
• Attempting to use brute force computation to do this potentially leads to many false positives
• the spectral responses of archaeological features are not unique
• the available data is very large.
• To use this data successfully requires a knowledge-led approach.
• This means a better understanding of how plants, land management, soil, weather, and the archaeology interact over time.
• Data mining of our benchmark data
• Help us ----- It’s open-data
Questions