Embed Size (px)
Citation preview
Friends and Partners in Aviation Weather (FPAW)July 12th 2017
Mike MatthewsJenny Colavito (FAA)
Alaska C&V Camera Imagery Analytics
Friends and Partners in Aviation Weather - 2Michael Matthews 12 July 2017
This material is based upon work supported by the Federal AviationAdministration under Air Force Contract No. FA8721-05-C-0002and/or FA8702-15-D-0001. Any opinions, findings, conclusions orrecommendations expressed in this material are those of theauthor(s) and do not necessarily reflect the views of the FederalAviation Administration.
© 2017 Massachusetts Institute of Technology.
Delivered to the U.S. Government with Unlimited Rights, as definedin DFARS Part 252.227-7013 or 7014 (Feb 2014). Notwithstanding anycopyright notice, U.S. Government rights in this work are defined byDFARS 252.227-7013 or DFARS 252.227-7014 as detailed above. Useof this work other than as specifically authorized by the U.S.Government may violate any copyrights that exist in this work.
Friends and Partners in Aviation Weather - 3Michael Matthews 12 July 2017
• DOT/FAA are increasing investment in cameras– Thousands of cameras installed nationwide– Primarily visual analysis– Difficult to monitor changing conditions
• Valuable weather condition information is not exploited– Ceiling and Visibility– Precipitation (type / intensity)– Pavement condition
(dry / wet / snow)– Fog
Video Analytics for Weather
Friends and Partners in Aviation Weather - 4Michael Matthews 12 July 2017
Lincoln Weather Video Analytics Timeline
1995 2000 2005 2010 2015 2020
Army National Highway FAA
Fund
ing
Bas
eIm
ager
y Lo
catio
n
Lexington, MA
North Adams, MA
Utah
Alaska
Lexington, MA
Friends and Partners in Aviation Weather - 5Michael Matthews 12 July 2017
FAA C&V Camera Analytics*
FAA – Alaska 260 Camera Sites
• Utilize camera imagery for real-time extraction of C&V variables– Stand-up Lincoln visibility algorithm in 2016– Tune algorithm for Alaska– Analyze performance with ASOS measurements – Research improved concepts and/or ceiling estimates
*Website is operated by an independent FAA office
ASOS = Automated Surface Observing System
Friends and Partners in Aviation Weather - 6Michael Matthews 12 July 2017
Extracting Visibility From Images
Edge DetectionImage Capture
10+ miles
¼ mile
Friends and Partners in Aviation Weather - 7Michael Matthews 12 July 2017
Visibility EstimatorNormalized Edge Extraction
Algorithm Flow
Edge DetectionImage Capture
Clear day composite image & edges
Image Registration
Visibility:10+
miles
Edge
Str
engt
h
Visibility
Friends and Partners in Aviation Weather - 8Michael Matthews 12 July 2017
Initial Alaska Camera Analysis Sites
Chandalar Shelf - NortheastShageluk - East
Tahneta Pass - WestHomer - Northeast
Friends and Partners in Aviation Weather - 9Michael Matthews 12 July 2017
Example Visibility Estimates:Tahneta Pass, March 14 to March 21, 2016March 14 March 15 March 17
March 18
March 16
March 19 March 21March 20
Edgestrength
Visibilityestimate
Visi
bilit
y (s
tatu
te m
iles)
Friends and Partners in Aviation Weather - 10Michael Matthews 12 July 2017
Example Co-Located Video / ASOSHomer, Alaska
NorthEast
SouthEastSouthWest
West
Five minute ASOS
ASOS = Automated Surface Observing System
Friends and Partners in Aviation Weather - 11Michael Matthews 12 July 2017
Example Preliminary ComparisonHomer, AK Northeast
ASOS Visibility (miles)
Vide
o Al
gorit
hm V
isib
ility
(mile
s)
<1
5-10
5-10
1-5
1-5
> 10
<1 > 10
12
222
1507
23657
16
3617
3
4
2 9
79
55 207
9629
Homer NE Camera, Jan 2014 – Dec 2015
Max Visibility = 12.0 SM
Overall match rate = 82%<10 mile match rate = 36%
Currently pursuing modifications to improve performance
Each cell = number of observations
Friends and Partners in Aviation Weather - 12Michael Matthews 12 July 2017
• FY2017 Activities– Initial Lincoln algorithm applied to AK images– Results are promising– Continuing to adapt to Alaskan environmental challenges
• Seasonal variations in composite• Sun angle limits• Camera positioning
– Performing comparison with ASOS (39 sites with 5-min ASOS)– Implementing prototype live capability (limited sites)
• FY2018 Activities– Engaging with Alaska Camera Program Office– Extending algorithm for sites with multiple cameras– Exploring expanded set of metrics and methods
• Sky brightness for visibility and/or sky cover• Machine learning techniques
Status and Future Work
