algorithm to process phenocam images eg · [email protected] figure 1. the algorithm...
TRANSCRIPT
![Page 1: Algorithm to Process Phenocam Images EG · sp.cortes1026@uniandes.edu.co Figure 1. The algorithm quantifies shifts in green and red from digital images . * Read image The algorithm](https://reader035.vdocuments.us/reader035/viewer/2022070900/5f3db7f144ae60751b533acf/html5/thumbnails/1.jpg)
Author: Leslie Goldman
© 2013 National Ecological Observatory Network, Inc. All rights reserved. The National Ecological Observatory Network is a project sponsored by the National Science Foundation and managed under cooperative agreement by NEON, Inc. This material is based upon work supported by the National Science Foundation under the following grants: EF-1029808, EF-1138160, EF-1150319 and DBI-0752017. Any opinions, findings, and conclusions or recommendations
expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.
Phenology is the study of the timing in plant and animal life
cycle events, which are mostly related to climate and
weather (Schwartz 2003)1. The study of plant phenology in
relation to global warming and increases in greenhouse
gases can give information on the fluctuations of plant
phenology due to climate change.
NEON will collect data for 30 years at 60 terrestrial sites in
different ecosystems, and this requires automatized and
standardized methods to gather reliable data. For this
project, an automated algorithm was designed to quantify
phenology in different ecosystems.
Introduction Algorithm Flowchart Modifications for NEON Determining Phenology Changes in the Long Term
The PhenoCam Network is a cooperative network that
archives and distributes imagery and derived data products
from digital Stardot cameras located at different research
sites around the world and North America. This network
archives image time series of vegetation that can be
analyzed with the PhenoCam Software Tools
(http://phenocam.sr.unh.edu/). NEON will use PhenoCam
network protocols and install Stardot cameras in 60
terrestrial sites.
Setup and Requirements
The PhenoCam Network
Excess of Green (EG) vs. DOY: Results of the Algorithm Applied to Different Ecosystems
Contact Information:
www.neoninc.org
Figure 1. The algorithm quantifies shifts in green and red from
digital images . *
Read image
The algorithm followed the PhenoCam Network tools by:
• Applying the suggested dark threshold (15%)
• Calculating the 90th quantile and averaging consecutive
images.
• Plotting the 90th quantile vs. day of the year.
• Calculating the day of the year with the image name.
The algorithm was modified to be automated and standardized:
• Region of interest is fixed and a grid is used for having
detailed phenology information per subarea.
• Calculation options are fixed: dark threshold and number of
consecutive images to average
• 6 Phenology metrics were calculated: green chromatic
coordinate (G/[G+R+B]), red chromatic coordinate
(R/[G+R+B]), normalized difference ratio ([G-R]/[G+R]),
simple band ratio (G/R), excess of green (2G-R-B), and
excess of red (2R-G-B).
• Algorithm calculates the DOY when the maximum, minimum
and inflection points occur.
Phenology metrics plots will be calculated for every year
during the 30 years of NEON’s data gathering. By having a
standardized algorithm, plots will be comparable between
different years and sites, so changes in phenology can be
measured.
Figure 2. Depiction of camera setup. *
Internship Project: Interpreting Canopy Phenology using an Automatic Image Analysis
Algorithm to Process Phenocam Images
Stephanie Cortés (Senior, Universidad de los Andes) Co-Author: Kevin Sacca (Junior, Rochester Institute of Technology) Mentors: Michael SanClements, Sarah Elmendorf (NEON)
Begin End
Time series images from PhenoCam
Dataset
Crop vegetation area
Split cropped area in color channels: R,
G,and B
Perform dark threshold of 15%
Calculate 6 phenology
metrics per pixel
15° below horizontal line
Image Composition:
80% Canopy, 20% sky
Camera connected to the
PhenoCam network
Images taken every 30
minutes (down to dusk)
with a resolution of 1.3 MP
Create a grid on the cropped area
Calculate 90th quantile of
phenometrics per subarea on grid
Average the 90th quantile for each
3 consecutive images
Plot 90th quantile vs. DOY
Curve fitting to a 6 degree
polynomial
Calculate minimum and
maximum points in the graph
Calculate inflection points
in the graph
Graphs and readout per
subarea of the grid with curve fit
and minimum, maximum, and
inflection points
Calculate day of the year (DOY) with the
image name
Figure 3. Possible climate change impacts on plant phenology.
Climate change could cause an earlier spring and a later autumn.
Bartlett Forest: Mixed Forest Wind River Forest: Coniferous Forest
Burns Sagebrush: Dryland Harvard Forest: Temperate Forest
0 100 200 300 4000
50
100
150
DOY
EG
0 100 200 300 4000
10
20
30
40
50
DOY
EG
0 50 100 150 200 250 300 350 4000
20
40
60
80
100
DOY
EG
0 50 100 150 200 250 300 350 40020
40
60
80
100
120
DOY
EG
The algorithm is standardized and automated to different ecosystems:
The figure shows the result plot with the 6 degree polynomial fit for four of the ecosystems tested. The results show a clear shift in the EG phenology metric (EG=2G-B-R) during spring and
summer, even though all of these ecosystems have different vegetation. These clear shifts show that the algorithm does not need changes depending on the ecosystem and that it can be
standardized and automated for the 60 terrestrial NEON sites.
Bartlett Forest during summer. * Wind River Forest during summer. *
Harvard Forest during summer. * Dryland during summer. *
--- 2011
--- 2041
Excess of Green vs. DOY
a b c d e
a. Winter
b. Spring
c. Summer
d. Early fall
e. Late Fall
1 Schwartz MD (2003) Phenology: an integrative environmental science. Kluwer, Dordrecht. * Images taken from the PhenoCam Network (http://phenocam.sr.unh.edu/).
Last image?
Figure 4. Algorithm output for 4 different ecosystems.
True
False