ai applications of earth observation data for disaster ...mind map of users. 2019.11.20 manila, the...
TRANSCRIPT
2019.11.20 Manila, the Philippines ASEAN IVO Forum 2019
AI Applications of Earth Observation Data for Disaster Monitoring in the Philippines
Roel M. de la CruzRocell Nino B. VicenteDOST-ASTI
2019.11.20 Manila, the Philippines ASEAN IVO Forum 2019
Background :
Project Title: Remote Sensing and Data Science (DATOS) Help Desk
Satellite earth observation (EO) supports data-driven assessment and decision-makingthrough the provision of timely information over a specific area of interest. This has becomeincreasingly vital in the four thematic areas of disaster risk reduction and management(DRRM) – prevention and mitigation, preparedness, response and rehabilitation and recovery– as rapid assessments can be done for urgent and prompt action.
To better provide analysis-ready data during critical events, the DOST-ASTI, through theRemote Sensing and Data Science Help Desk, or the DATOS Project, is leveraging ArtificialIntelligence to develop applications for DRRM using satellite images along with other criticalinfrastructure such as the Philippine Earth Data Resource and Observation (PEDRO) Center’ssatellite ground receiving station and the Computing and Archiving Research Environment’s(COARE) high-performance computing facility.
2019.11.20 Manila, the Philippines ASEAN IVO Forum 2019
Background :
The Remote Sensing and Data Science (DATOS) Help Desk of the Department of Science andTechnology-Advanced Science and Technology Institute (DOST-ASTI) uses the agency’s HighPerformance Computing (HPC) facility in processing satellite images for various disaster-related and agricultural mapping outputs.
DOST-ASTI acquires terabytes of satellite images from several commercial optical and radarsatellite subscriptions through the PEDRO Center Ground Receiving Station, which are storedand processed in the Computing and Archiving Research Environment (COARE) facility. DATOSthen applies Artificial Intelligence (AI) or Time-Series Analysis to detect and map out featuresfrom these satellite images.
Project Title: Remote Sensing and Data Science (DATOS) Help Desk
2019.11.20 Manila, the Philippines ASEAN IVO Forum 2019
DATASETS
2019.11.20 Manila, the Philippines ASEAN IVO Forum 2019
HPC
Object Detection
(GPU)
Time-Series
Analysis
(CPU)
Earth
Observation Data• Maps
• Analysis
METHODOLOGY
2019.11.20 Manila, the Philippines ASEAN IVO Forum 2019
OUTPUTS
2019.11.20 Manila, the Philippines ASEAN IVO Forum 2019
AI MODELS
Synthetic Aperture Radar
15-m resolution
Flood maps
Fish pens
Fish ponds
Ships
Landsat Images30-m resolution
Built-up Areas
Water Bodies
Vegetation
Clouds/Shadows
Bare Soil
Planet Imagery3-m resolution
Built-up Areas
Water Bodies
Vegetation
Clouds/Shadows
Bare Soil
Irrigation Network
Road Network
VHR Images0.3-m resolution
Mango Trees
Coconut TreesPili Trees
Damaged BuildingsBuildings
2019.11.20 Manila, the Philippines ASEAN IVO Forum 2019
.
DBM-DIME*
DOT
DPWH
NDRRMC
PHIVOLCS
DSWD
Coast Watch
AFPPCA
SRA
BPSU
DA-FPOPD
NAMRIA
DENR
BFAR
Sugarcane Mapping
Mango Tree Mapping
Corn Mapping
Port Detection
Building Detection
Damaged Buildings
Detection
Drought Maps
Flood Maps
Earthquake Interferogram
Critical Infrastructure Mapping
Land Cover Classification
Ship Detection
Road Detection
Bridge Detection
Change Detection
Coconut Mapping
Rice Mapping
Fish Pen Detection
Agencies that have MOA on Research Collaboration with DOST-ASTI through DATOS
Legend:
Agencies that are recipients of DATOS outputs as needed
AI for EO
Infra
DRR-CCA
Security and
Defense
Food Security
Monitoring
Agencies with ongoing talks and/or possible users*MOA is through DOST Main Office
MIND MAP OF USERS
2019.11.20 Manila, the Philippines ASEAN IVO Forum 2019
SUMMARY
The DATOS Project developed techniques to extract features from remotely-sensed data through automation of remote-sensing workflows and the use of Neural Networks. Examples are automated flood detection from multi-temporal SAR images, land cover detection, road network detection, mango tree detection and agricultural plantation mapping.
Moving forward, DOST-ASTI is looking to further expand the reach and adoption of its space-based tools and applications through a combination of stakeholders’ meetings and trainings. These initiatives seek to capacitate other government agencies to be able to ingest, process, and analyze earth observation data, creating a community of users for better incidence impact to the Filipino people.