DATOS: Artificial Intelligence in Remote Sensing – PRESS RELEASE

22 Feb 2019 5:27 PM

The Remote Sensing and Data Science: DATOS Help Desk (or the DATOS Project), funded by DOST-PCIEERD, is the DOST-ASTI’s geospatial applications initiative that applies Artificial Intelligence (AI), Machine Learning, and other data science techniques to remotely sensed data.

DOST-ASTI currently has an increasingly extensive collection of data and infrastructure for data science. The agency is home to the Philippine Earth Data Resources Observation (PEDRO) Center, the ground receiving station for images from the Diwata micro-satellites and other commercial satellites that the DOST-ASTI is subscribed to. The agency also deployed over 2,000 automated hydro-meteorological sensors that collect near real time weather and water level data from all over the Philippines.

These satellite images and weather data are archived and are processed via the Computing and Archiving Research Environment (COARE) Facility. COARE is a High-Performance Computing and Cloud facility that allows free access of its services to students, researchers, and data analysts. Akin to a super computer, the facility has thousands of cores of CPU and GPU compute power, several petabytes of storage, and a network speed on 10 gigabytes per second to serve as a platform for easy storage, analysis, and sharing of environmental and geospatial data. DOST-ASTI is also home to PREGINET, the country’s only research and education network (REN) that interconnects academic, research, and government institutions and international RENs, like a high-speed internet connection.

All these data, on top of data produced by previous DOST projects, and facilities are used by the DATOS Project to produce geospatial outputs that can be used for disasters, agriculture, and other purposes. The DATOS Project implements Remote Sensing (RS), Geographic Information System (GIS), Machine Learning, AI, and other data science techniques to produce these geospatial outputs.

As an example, the DATOS Project has developed a way to map out crops by using satellite images and extracting the crops’ “temporal signature”, which is similar to a voice recognition technique (where different words of varying lengths can be identified as the same word) determined via radar satellite images. The project has a standing partnership with the Sugar Regulatory Administration (SRA) to help automate the mapping of sugarcane plantations for yield prediction and disaster monitoring. The same methodology can be used to map out other crops with known temporal signature such as rice and corn.

DATOS also produces flood situation maps by getting satellite images and letting an AI identify flooded areas from these imageries. In the event of severe weather disturbances, DATOS detects floods in areas that are hit by heavy rainfall and sends these mapped out areas to the respective DOST Regional Offices. The team also uses AI to detect objects from satellite images, such as mango trees as part of a current partnership with the Bataan Peninsula State University to help automate the latter’s mapping of their province’s mango trees. Other objects that have been detected by the project from satellite images using AI include road networks, ships, land cover classes, and built-up areas.

For more content, follow the DATOS Project on Facebook.

Established in 1987 by Executive Order No. 128, the Advanced Science and Technology Institute (ASTI) is an attached agency of the Department of Science and Technology (DOST) that undertakes scientific research and development and technology transfer in the advanced fields of Information and Communication Technology (ICT), computing, electronics and their applications.

The DOST-ASTI continues to dedicate itself to developing and delivering technology solutions to enable a productive, globally competitive and resilient Filipino society.

Media Enquiries
Jo Brianne Briones
Information Officer III
+632 2498500



Image 1: Flood Situation Map in Northern Leyte on 30 January 2019, 6:00 PM (date of acquisition of satellite imagery). Areas in red are floods detected by AI from Sentinel-1 satellite images. Map produced by the DATOS Project.

Image 2: Map of detected sugarcane plantations in an area in Tarlac from satellite images using Dynamic Time Warping in. Cyan areas in the image are detected sugar plantations, while yellow boxes are sugarcane plantation digitized from optical satellite images and ground-validated by the Sugar Regulatory Administration. Map produced by the DATOS Project.

Image 3: Left - Digital Globe satellite image of an area in Bataan where mango trees are planted. Right - Mango trees (in blue) detected by the DATOS Project from the satellite image. The AI model was trained using Digital Globe images with a pixel resolution of 3 meters. Red, Green, Blue, and NIR bands were used as input parameters to the AI Neural Network. Three indices (NDVI, NDCI, NDWI) were also derived from the satellite bands and included as inputs to the AI. The AI architecture used was U-NET, a Convolutional Neural Network specifically designed for image segmentation. Integrating an Image Detection framework (YOLO v3) to the existing model is currently in development to be able to separate and count the number of segmented mango trees.