ARSF Data ProcessingConsequences of the Airborne Processing Library
Mark Warren
Plymouth Marine Laboratory, Plymouth, UK
RSPSoc 2012 – Greenwich, London
RSPSoc 2012 – Greenwich
Overview
• Airborne Research and Survey Facility (ARSF)– Who are we, what do we do
• Airborne Processing Library (APL)– Hyperspectral processing suite
• Geocorrection– Airborne hyperspectral images
• Potential error sources– Mapping using APL
RSPSoc 2012 – Greenwich
ARSF: Who are we
• Airborne Research and Survey Facility (ARSF)• NERC facility • Supporting UK & European science
– Dornier 228 aircraft• Two hyperspectral sensors
• Full waveform LiDAR
• Medium format digital camera
• Plymouth / Gloucester
RSPSoc 2012 – Greenwich
Hyperspectral Remote Sensing @ ARSF
• Specim Eagle sensor– Visible & Near Infra-Red
• 400nm - 1000nm
– 'Push-broom' sensor– Field of view ~37 degrees
• Specim Hawk sensor– Short Wave Infra-Red
• 1000nm – 2500nm
– 'Push-broom' sensor– Field of view ~24 degrees
RSPSoc 2012 – Greenwich
Example data – Poole UK
Left: Eagle Right: Hawk
RSPSoc 2012 – Greenwich
Airborne Processing Library (APL)
• Software suite developed to process ARSF hyperspectal data– Radiometric calibration– Geocorrection
• Cross purpose – in-house + end user– Windows, Linux– Graphical User Interface or Command Line
RSPSoc 2012 – Greenwich
Point of View of ARSF user
• ARSF data delivered at “level 1”– Radiometric calibration– Navigation synchronisation– [2012 onwards also delivered mapped]
• User can apply additional algorithms– e.g. Atmospheric correction
• User can geocorrect the data with APL– Produce maps of data
RSPSoc 2012 – Greenwich
Unmapped data
• Little Rissington Airfield
• Difficult to find targets– Distortions– Direction of flight– No fixed X,Y coordinates
• Geocorrection can help
RSPSoc 2012 – Greenwich
Geocorrection – What?
• What is it?– Associating position information– Mapping to a real-world projection
• Benefits of geocorrecting / mapping– Easier to identify targets– Compare data from other map sources
• Limitations of geocorrecting / mapping– Can introduce different distortions– Can give misleading results
RSPSoc 2012 – Greenwich
Geocorrected / Mapped data
RSPSoc 2012 – Greenwich
Geocorrection – How?
• Stage 1 – create the mapping– Position / attitude / sensor pixel vectors– Per-pixel position information
• Stage 2 – resample data– Output pixel size– Interpolation– Fill the mapped grid using stage 1 mapping
RSPSoc 2012 – Greenwich
Geocorrection Limitations for Airborne Data
• Airborne RS data usually localised areas– Projection internal distortion not big issue
• Platform stability– Wind / atmospheric buffeting – Roll / pitch / yaw
• Position accuracy– GPS constellation + ground stations
• Sensor – Stability of sensor head (internal movements)– Lens distortions
RSPSoc 2012 – Greenwich
Potential Error Sources – In the Data
• Level 1 data– Navigation
• Position accuracy – lateral shift
• Synchronisation – distortions and shifts
RSPSoc 2012 – Greenwich
Zoom – synchronisation error
RSPSoc 2012 – Greenwich
Potential Error Sources – In the Data
• Auxiliary data– Digital Elevation Model – per-pixel positional errors– More accurate DEM the better
RSPSoc 2012 – Greenwich
Potential limiting sources – Mapping 1
• Pixel size– Try and stay similar to spatial resolution
• Related to aircraft height above surface
– Size effects• Too small - repeated data (not more data!)
• Too large - lost data
• 'Blocky' image
RSPSoc 2012 – Greenwich
Pixel size
• 3 images at the same zoom level– 10m pixel – shows lost data– 2m pixel– 0.5m pixel – shows repeated data
RSPSoc 2012 – Greenwich
Potential limiting sources – Mapping 2
• Interpolation– Required for transformation from 1 grid to another– Nearest neighbour
• Guarantees 'real' observed values
• 'blocky' image
– Bilinear / Bicubic• Unobserved (maybe unrealistic) values
• Smoothed data, visually pleasing image
• Problems with in-situ data comparisons
RSPSoc 2012 – Greenwich
Interpolation
• Nearest Neighbour vs Cubic
RSPSoc 2012 – Greenwich
Atmospheric Correction
• Atmospheric Correction– Level 1 vs Mapped geometries– More spectral coverage the better
• Problem: separate Eagle / Hawk– Combine the spectra– Problems
• Different spatial resolution
• Different look vectors
• Different swath widths
– Partial geocorrection of both and combine nearest points
RSPSoc 2012 – Greenwich
Summary
• Intro to ARSF hyperspectral instruments• Problems associated with geocorrecting RS
data• Potential error / limiting effects• Future atmospheric correction products
RSPSoc 2012 – Greenwich
Thank you for listening
• Any questions?
RSPSoc 2012 – Greenwich
RSPSoc 2012 – Greenwich
Potential limiting sources – Mapping 3
• Multiple bands and Masking– Masking data
• Insert a “null” value
• Interpolated over
– Multiple bands• Spectral analysis
– incorrect profile if some bands masked
• Assumes sensor view vectors same for each band