Early Warning Disaster System based on Mu

BLUE VISION
EARLY WARNING DISASTER SYSTEM
BASED ON MULTISPECTRAL REMOTE
SENSING OF THE EARTH
Plamenka Borovska, Desislava Ivanova
Computer Systems Department
Technical University of Sofia
CONTENT
Remote Sensing of the Earth
 Early Warning Disaster System – Problem
Statement
 Blue Vision Software
 Experimental Platform
 Results and Analysis
 Conclusion and Future work

REMOTE SENSING


OF THE
EARTH
Earth
remote
sensing
is
data
collection
on
the environment, geology, climate, and other characteristics
of the Earth by means of sensors positioned in the air or in
Earth orbit.
Sensors can be mounted on different platform
ladder, tall building, crane
 balloon
 plane or other airborne structure


satellite orbiting the Earth
REMOTE SENSING

EARTH
Sensors have different parameters in terms of
resolution:

Spectral Resolution



from 3 to 220 spectral bands
Spatial Resolution

from 0.4 to 1000 m
Temporal Resolution


OF THE
from few hours to more than 16 days
Remote sensing is quick and relatively cheap
technique for monitoring Earth’s surface on a large
scale
EARLY WARNING DISASTER SYSTEM
PROBLEM STATEMENT

Disasters can be detected





Fires
Deforestation
Soil Salinity
Water pollution
Flood detection
Water pollution
Soil Salinity
Deforestation
Fire
Flooding
BLUE VISION SOFTWARE

BLUE VISION Software:
parallel computational models
 5 modules
 implement parallel algorithms for detection of:

PARFIRE for fire detection
 PARDEFOR for detection of deforestation
 PARSOIL for detection of high soil salinity
 PARWATER for detection of water pollution
 PARFLOOD for flood detection

MODIS - MODERATE RESOLUTION IMAGING
SPECTRORADIOMETER

On board of NASA’s Terra and Aqua satellites

near-polar orbit


four images per day


two day and two night time
TERRA
satellite
36 spectral bands


705 km above the Earth surface
in visible and infrared spectrum
spatial resolution



250 m (bands 1-2)
500 m (bands 3-7)
1000 m (bands 8-36)
AQUA
satellite
HDF (Hierarchical Data Format) files
 Provides relevant data for disaster monitoring
based on multispectral analysis

ALGORITHMS


Based on multispectral analysis
Utilize data in several spectral bands

NASA’s Enhanced Fire Detection Algorithm


Normalized Difference Vegetation Index (NDVI)







BRDF – Bidirectional Reflectance Distribution Function
Albedo
Vegetation indices
FPAR - Fraction of Photosynthetically Active Radiation
LAI – leaf area index
Soil Salinity Index

NDVI - normalized difference vegetation index

Salinity Index
NASA’s Chlorophyll Water Analyses Algorithm


the algorithm uses the temperatures generated in the thermal infrared
spectral channels of MODIS sensor at a wavelength of 4 μm and 11μm.
The algorithm determines the areas of water pollution by assessing the
concentration of chlorophyll, which is used as an indicator of
phytoplankton biomass.
Time change detection
PARALLEL COMPUTATIONAL MODELS

SPMD parallel paradigm


small communication overhead
Results gathered by master process
 position coordinates
PARALLEL COMPUTATIONAL MODELS

Message Passing Interface (MPI)
distributed memory model
 course grained parallelism


Multithreading (OpenMP)
shared memory model
 fine grained parallelism

PARALLEL COMPUTATIONAL MODELS

Hybrid parallel model

MPI + OpenMP
HDF
EXPERIMENTAL PLATFORM

Bulgarian Supercomputer –
IBM Blue Gene/P

Heterogeneous Compact
Computer Cluster –
“High-Performance and Cloud
Computing Lab”, Computer System
Department, Technical University - Sofia
“High-Performance and Cloud
Computing Lab”,
Computer System Department
RESULTS AND ANALYSIS


Linear Speedup
Good Scalability
 Processing
of one
Data Set
RESULTS AND ANALYSIS
 Results
Visualization
 Google
map
RESULTS AND ANALYSIS


Multispectral Analysis of the Remote Sensing of
the Earth
BLUE VISION SOFTWARE


PARFIRE, PARDEFOR, PARSOIL, PARWATER,
PARFLOOD
Good Scalability with respect to data sets and
parallel computer architectures
FUTURE WORK

Improvements and Extension of Existing Early
Warning Disaster System based on Multispectral
Remote Sensing of the Earth

RADAR SATELLITE IMAGING


Earth-observing radar satellites data - Radar satellite data
offer the great advantage, over its optical counterparts, of not
being affected by meteorological conditions such as clouds,
fog, etc., making it the sensor of choice when continuity of
data must be ensured.
CLASSIFIER - Improving radar observations of disasters with
classifier based on parallel algorithm for implementation of Fast
Wavelet Transforms (Mallat's pyramid algorith)
THANK YOU FOR YOUR ATTENTION!